CN103905822A - Low-bit compressed image deblocking method based on visual sensitivity and spatial filtering - Google Patents
Low-bit compressed image deblocking method based on visual sensitivity and spatial filtering Download PDFInfo
- Publication number
- CN103905822A CN103905822A CN201410112243.4A CN201410112243A CN103905822A CN 103905822 A CN103905822 A CN 103905822A CN 201410112243 A CN201410112243 A CN 201410112243A CN 103905822 A CN103905822 A CN 103905822A
- Authority
- CN
- China
- Prior art keywords
- block
- image
- blocking effect
- filtering
- texture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention relates to a low-bit compressed image deblocking method based on visual sensitivity and spatial filtering. The method includes the steps that an input distorted image is divided into a flat region, an edge region and a texture region according to visual sensitivity characteristics; then a Sobel operator is selected to extract gradient information of pixels of the image, a threshold value is set to enable the image to be divided into an edge region, a texture region and a flat region; the image is further divided into flat blocks, edge blocks and texture blocks; filtering is conducted on the blocks in different filtering modes, and finally block information obtained after filtering is integrated to obtain a processed image. According to the method, the human eye visual sensitivity characteristics are applied to classification of compressed distorted image texture, and the blocking effect in the image is subdivided into six types according to the neighboring relation of the texture blocks. The classification method is simple and effective, an appropriate filter can be selected specifically and easily to perform filtering processing on the blocking effect.
Description
Technical field
The present invention relates to image processing method technical field, relate in particular to a kind of low bit compression image deblocking effect method based on visual sensitivity and airspace filter.
Background technology
Digital multimedia technical development is at present rapid, and in order to meet digital picture in transmission, the demand in the processes such as storage, can compress processing to image conventionally.Compression and coding standard based on block DCT transform as JPEG, MPEG, H.261, H.263 etc. be widely used in the various aspects of multimedia technology.But this coded system based on block DCT transform is in the time of low bit rate, image after decompress(ion) can produce serious blocking effect distortion, and the lower blocking effect of the bit rate of image is more obvious, blocking effect distortion has caused and has had a strong impact on the subjective and objective quality of image, cause image not only cannot meet people's subjective demand, also can be to the Intelligent Recognition of image, the work such as detection cause important obstruction.People often adopt the method for reprocessing to reduce blocking effect.These class methods can be divided into two classes: the post-processing technology of recovering based on image and the post-processing technology based on figure image intensifying.The post-processing approach recovering based on image needs certain priori such as decoded information and distortion model conventionally, and the general amount of calculation of these class methods is larger, is difficult to real-time processing.Post-processing approach based on figure image intensifying does not rely on any decoded information, can independently remove blocking effect, has obtained research widely.In the post-processing approach based on strengthening, airspace filter is a kind of the most basic method, and the method is directly processed the brightness value of pixel in image, the feature that has flexibly, intuitively, easily realizes.
A kind of the simplest airspace filter method is to adopt simple space to move constant low pass filter, and all pixels in image are carried out to filtering.Although the method has been removed blocking effect, destroy detail of the high frequency abundant in image, can make the filtered entire image excessively fuzzy.Because the blocking effect in image is mainly because quantization error causes, the variation of picture material can cause the different expression form of the blocking effect of compressed images, image zones of different is selected to the filter of different level and smooth intensity, has very important meaning.Therefore airspace filter method adopts the method for territorial classification conventionally, and image is divided into different regions, more targetedly zones of different is carried out to filtering processing, can not make image excessively fuzzy so on the one hand, can protect the detail of the high frequency of image again.Usual way is to adopt high strength level and smooth at flat site, because flat site does not exist high-frequency information, the high-intensity image that smoothly can not make is excessively fuzzy.For non-flat forms piece, owing to there being more high-frequency information, application low-intensity smothing filtering can be preserved detail of the high frequency preferably.But in the method, the region unification that contains limbus information is classified as texture region, does not process separately.Experimental result shows, is subject to the impact of quantizing distortion, and larger impact can be caused to the subjective quality of image in this class region.Therefore, adopt the filter identical with texture region can not process well the quantizing distortion problem of fringe region.
In recent years, researcher is studied the visual sensitivity feature of human eye, and result of study shows, human eye has different susceptibilitys to the blocking effect distortion in different texture region.Under different code checks, human eye is found the susceptibility of blocking effect distortion after tested: eye-observation mainly concentrates on to the obvious place of blocking effect the low frequency flat site that luminance transformation is less; Except flat region, the quantizing distortion phenomenon of marginal portion is also obvious; And at high frequency texture region, because the directivity of texture is many and complex structure, make that the blocking effect in this part region is more difficult to be discovered by human eye.This phenomenon just human visual system (HVS) is covered characteristic, and human eye has for different texture region that different susceptibilitys causes.So be necessary to utilize this characteristic that a kind of low bit compression image deblocking effect method is set.
Summary of the invention
The object of the invention is, towards low bit rate jpeg compressed image, adopt a kind of deblocking effect method based on visual sensitivity and airspace filter combination, solve the blocking effect problem of dtmf distortion DTMF in low bit compression image.
For achieving the above object, the technical scheme that the present invention deals with problems is: a kind of low bit compression image deblocking effect method based on visual sensitivity and airspace filter is provided, it is characterized in that, described method comprises the steps:
S1: the distorted image of input is divided into flat region, marginal zone and texture area according to visual sensitivity characteristic;
S2: select Sobel Operator to extract the gradient information of image slices vegetarian refreshments, entire image is being carried out to after Grad asks for, adopt three kinds of Component Models, setting threshold, is divided into marginal zone, texture area and flat site by image;
S3: respectively according to 8 × 8 big or small piecemeals, distorted image region is just further divided into: flat block, edge block and texture block by three regions of distorted image;
S4: again according to the neighbouring relations of the type of region unit and piece and piece, blocking effect type in distorted image is divided into: the blocking effect between the blocking effect between blocking effect, edge block between flat block, blocking effect, flat block and the edge block between texture block, the blocking effect between the blocking effect between edge block and texture block and texture block and flat block;
S5: for the blocking effect between flat block, adopt the loop block-eliminating effect filtering algorithm in video compression standard H.264, remove the blocking effect between flat block with the filter of filtering strength maximum, only selecting adjacent block in the time of filtering is all the border of flat block;
S6: for the blocking effect between edge block, adopt a kind of self-adaptive smooth filtering mode to carry out filtering, first described filtering assessed the block effect intensity of the every a line centre of block boundary pixel, then selects the smooth function of an applicable adaptive control by block effect intensity control to carry out filtering to blocking effect;
S7: the blocking effect between edge blocks and flat block and between edge block and texture block, in order not make entire image too fuzzy because filtering causes, does not process for this class blocking effect;
S8: for the border between the blocking effect between texture block and texture block and flat block, adopt the similar simplification loop block-eliminating effect filtering method of situation of the blocking effect between flat block;
Further, in described step S2, select Sobel Operator to extract the gradient information of image slices vegetarian refreshments, adopt four direction template, be respectively horizontal direction, vertical direction, the template of 45 degree directions and 135 degree directions, for piece image, first carry out convolution with this image respectively by this four direction template, obtain the Grad on image four direction, and select in four Grad maximum one, be denoted as gmax; Then adopt three kinds of Component Models that image is divided into fringe region, texture region and flat site; Body is suddenly:
(1), calculated threshold, TH1=γ × gmax and TH2=δ × gmax, TH1, TH2 are height threshold value, as the foundation of differentiating pixel Texture classification, γ span is between 0.1～0.2, δ span is between 0.01～0.1;
(2), adopt following formula that image pixel is divided into edge, texture or flat site;
Wherein, G (i, j) is the Grad of each pixel in image, and in the time that the Grad of pixel (i, j) position is greater than TH1, this pixel is just divided into marginal zone; In the time that the Grad of pixel (i, j) position is less than TH2, this pixel is just divided into flat region; If the Grad of pixel (i, j) position between TH1 and TH2 time this pixel be just divided into texture area.
Further, describedly by three regions of distorted image according to the concrete sorting technique of 8 × 8 big or small piecemeals be respectively: suppose a, b is respectively the flat block of two 8 × 8 pixel sizes, and a, the position relationship of b is that left and right is adjacent, and the right side of a 4 is listed as with a left side 4 row pixels of b and forms an image block C, and the blocking effect in the middle of a, b will be intactly retained in C, therefore to a, between b, the filtering of blocking effect is that C is carried out to filtering operation actually; Work as a, when the position relationship of b is neighbouring, form image block Cv; If a, b is flat block, and the blocking effect in C is the blocking effect between flat block, if a, b one of them be flat block another be edge block, the blocking effect in C is the blocking effect between flat block and edge block, can obtain by that analogy 6 kinds of blocking effect classifications.
Further, the method for described composition image block C and Cv, establishing piece C is the new piece of adjacent block a and b composition, establishes any a line 8 pixels from left to right in C and is expressed as successively: a4, a3, a2, a1, b1, b2, b3, b4; In C, with the row indicating in black dotted lines, adopt formula to carry out below filtering to any a line in C so:
a
1′=(a
3+2a
2+2a
1+2b
1+b
2+4)/8
a
2′=(a
3+a
2+a
1+b
1+2)/4
a
3′=(2a
4+3a
3+a
2+a
1+b
1+4)/8
A in formula
1', a
2' and a
3' be a1, a2, the result of a3 after after filtering, a4 point is not processed; When b point value is carried out to filtering, only need in filter, change a point pixel of relevant position in formula into b point pixel, b point pixel changes a point pixel into; Be the filtering mode that b1 is adopted to a1, b2 adopts the filtering mode of a2, and b3 adopts the filtering mode of a3, and b4 adopts the filtering mode of a4 to carry out filtering; Working as a, when the position of b is neighbouring, is Cv, only need to be by a, and b is 90-degree rotation simultaneously, and Cv just becomes the form the same with C, and when then adjacent according to left and right, C situation is carried out filtering.
Further, be all the situation of edge block for described adjacent block, the building form of continuing to use equally described C piece obtains C piece, if the pixel intensity in C centre position has obvious saltus step, this blocking effect of C is simulated with a two-dimentional step function blk, shown in following formula:
In formula, numerical value 1/2 and-1/2 has represented the alias in the middle of image block C; First the block effect intensity of the every a line centre pixel in C piece is assessed, following formula, uses c (i, j) to represent the pixel value that i is capable, j is listed as in C here.
β=[c(i,6)-3×c(i,5)+3×c(i,4)-c(i,3)]/2
In formula, β represents the block effect intensity in C piece centre position; Then select an applicable smooth function to replace the step function that produces blocking effect, the detailed information of considering edge block is more, and there is the position texture structure complexity of blocking effect, therefore only boundary pixel carried out slightly smoothly the following formula of smooth function of employing:
In formula, β level obtains according to block effect intensity β, can control adaptively the shape of smooth function, when β level value more hour smooth function approach original step function; Smooth function can change border pixel values in Min. ground, reduces the brightness saltus step on image block border, can not cause the fuzzy of image; In the time processing, this smooth function is carried out to discretization, construct one dimension smooth function, shown in following formula:
de_blk(j)=[f(-49)f(-35)f(-21)f(-7)f(7)f(21)f(35)f(49)]
c'(i,j)=c(i,j)+β×[de_blk(j)-blk(j)]
Here, de_blk (j) is the array by 1 × 8 size of f (x) generation, c'(i, j) be the capable result after filtered of the i of C, here c (i, j) refer to the capable all elements of i of C, blk(j) refer in C that a line is modeled into the value after blk arbitrarily, each in this sampled images is passed through after level and smooth can effectively remove blocking effect.
Further, to the blocking effect between described edge block and flat block and between edge block and texture block, in order not make entire image too fuzzy because filtering causes, this two classes blocking effect is not processed.
Further, to the border between the blocking effect between described texture block and texture block and flat block, adopt the similar simplification loop block-eliminating effect filtering method of situation of the blocking effect between flat block, when difference is filtering, the pixel of a3 position and b3 position is not carried out to filtering processing.
Compared with prior art, the invention has the beneficial effects as follows: the method has been utilized human eye vision sensitivity characteristic, and distorted image is divided into flat block edge block and texture block.And further the blocking effect type in image is subdivided into: the blocking effect between the blocking effect between blocking effect, edge block between flat block, the blocking effect between texture block, edge block flat block, the blocking effect between the blocking effect between edge block and texture block and texture block and flat block; Then combine airspace filter method, adopt the filtering algorithm of suitable varying strength, remove pointedly blocking effect, improve the weak point of existing algorithm to the edge class texture filtering in image, solve the problem that goes out there will be new distortion in existing method after filtering at edge class texture, in this simultaneously, this method is also better than existing method to a certain extent to the block effect filtering effect of flat site, and can retain preferably the detail of the high frequency in original image.Compared with the conventional method, the filtered image that the inventive method obtains has higher subjective and objective quality.Feature of the present invention can be summarized as following 3 points:
1. human eye vision sensitivity characteristic is applied in the classification of compression artefacts image texture, and according to the neighbouring relations of texture block, the blocking effect in image is segmented for 6 classifications.Sorting technique is simply effective, is beneficial to and selects more targetedly suitable filter to carry out filtering processing to blocking effect
2. in Texture classification process, Edge texture and common texture are made a distinction, efficiently solve existing method and process due to human eye is considered as to common texture blocking effect than the blocking effect at more sensitive Edge texture place the new distortion phenomenon causing.
3. utilize human eye vision susceptibility intensity to select the airspace filter method that intensity is suitable, different classes of blocking effect is carried out to filtering targetedly, not only strengthened the filter effect of flat site, also effectively protected the detail of the high frequency in source images simultaneously.
Brief description of the drawings
Fig. 1 is the general frame of the inventive method;
Fig. 2 is the system flow chart of the inventive method;
Fig. 3 is blocking effect schematic diagram of the present invention and needs filtered level and vertical boundary schematic diagram;
Fig. 4 is the inventive method and existing methodical subjective quality comparative effectiveness figure:
In Fig. 4: (a) be original image; (b) be the image (PQ=10, bpp=0.36) after JPEG compression;
(c) be existing methodical design sketch; (d) be design sketch of the present invention;
Fig. 5 is the raising curve chart of the inventive method to compressed image objective quality.
Embodiment
As shown in Fig. 1～Fig. 5, a kind of low bit compression image deblocking effect method based on visual sensitivity and airspace filter of the present invention, the method comprises the steps:
The first step: the distorted image of input is divided into flat region, marginal zone and texture area according to visual sensitivity characteristic;
Second step: select Sobel Operator to extract the gradient information of image slices vegetarian refreshments, entire image is being carried out to after Grad asks for, adopt three kinds of Component Models, setting threshold, is divided into marginal zone, texture area and flat site by image;
The 3rd step: respectively according to 8 × 8 big or small piecemeals, distorted image region is just further divided into: flat block, edge block and texture block by three regions of distorted image; Block type is determined by the corresponding at most texture type of texture pixel in this piece;
The 4th step: again according to the neighbouring relations of the type of region unit and piece and piece, blocking effect type in image is divided into: the blocking effect (F|E) between the blocking effect (E|E) between blocking effect (F|F), edge block between flat block, blocking effect (T|T), flat block and the edge block between texture block, the blocking effect (T|F) between blocking effect (E|T) and texture block and flat block between edge block and texture block;
The 5th step: for the blocking effect between flat block, adopt the loop block-eliminating effect filtering algorithm in video compression standard H.264, remove the blocking effect between flat block with the filter of filtering strength maximum, in the time of filtering, only selecting adjacent block is all the border of flat block;
The 6th step: for the blocking effect between edge block, adopt a kind of self-adaptive smooth filtering mode to carry out filtering, first described filtering assessed the block effect intensity of the every a line centre of block boundary pixel, then selects the smooth function of an applicable adaptive control by block effect intensity control to carry out filtering to blocking effect;
The 7th step: the blocking effect between edge blocks and flat block and between edge block and texture block, in order not make entire image too fuzzy because filtering causes, does not process for this class blocking effect;
The 8th step: for the border between the blocking effect between texture block and texture block and flat block, adopt the similar simplification loop block-eliminating effect filtering method of situation of the blocking effect between flat block;
Execution mode is further in the present invention, in described second step, select Sobel Operator to extract the gradient information of image slices vegetarian refreshments, adopt four direction template, be respectively horizontal direction, vertical direction, the template of 45 degree directions and 135 degree directions, for piece image, first carry out convolution with this image respectively by this four direction template, obtain the Grad on image four direction, and select in four Grad maximum one, be denoted as gmax; Then adopt three kinds of Component Models that image is divided into fringe region, texture region and flat site; Specifically suddenly:
(1), calculated threshold, TH1=γ × gmax and TH2=δ × gmax, TH1, TH2 are height threshold value, as the foundation of differentiating pixel Texture classification, γ span is between 0.1～0.2, δ span is between 0.01～0.1;
(2), adopt formula (1) that image pixel is divided into edge, texture or flat site;
Wherein, G (i, j) is the Grad of each pixel in image, and in the time that the Grad of pixel (i, j) position is greater than TH1, this pixel is just divided into marginal zone (edgepixel); In the time that the Grad of pixel (i, j) position is less than TH2, this pixel is just divided into flat region (smoothpixel); If the Grad of pixel (i, j) position between TH1 and TH2 time this pixel be just divided into texture area.
(2) by three regions of the above image respectively according to 8 × 8 big or small piecemeals, image-region is just further divided into: flat block (flat block, F), edge block (edge block, E) and texture block (texture block, T).As shown in classification results schematic diagram in Fig. 1,2.
Execution mode is further in the present invention, describedly by three regions of distorted image according to the concrete sorting technique of 8 × 8 big or small piecemeals be respectively: according to the neighbouring relations of the type of region unit and piece and piece, the blocking effect type in image is divided into: the blocking effect between the blocking effect between blocking effect, edge block between flat block, the blocking effect between texture block, edge block and other type blocks and the blocking effect between texture block and flat block.Suppose a, b is respectively the flat block of two 8 × 8 pixel sizes, and a, the position relationship of b is that left and right is adjacent, as shown in Figure 3, the right side of a 4 is listed as with a left side 4 row pixels of b and forms a new image block C, the blocking effect in the middle of a, b will be intactly retained in C, therefore to a, between b, the filtering of blocking effect is that C is carried out to filtering operation actually; Work as a, when the position relationship of b is neighbouring, form new image block as Fig. 3 terminal Cv; If a, b is flat block, and the blocking effect in C is the blocking effect between flat block, if a, b one of them be flat block another be edge block, the blocking effect in C is the blocking effect between flat block and edge block, can obtain by that analogy 6 kinds of blocking effect classifications.
Execution mode is further in the present invention, the method for described composition image block C and Cv, and establishing piece C is the new piece of adjacent block a and b composition, establishes any a line 8 pixels from left to right in C and is expressed as successively: a4, a3, a2, a1, b1, b2, b3, b4; In C, with the row indicating in black dotted lines, adopt formula (2), (3), (4) to carry out filtering to any a line in C so:
a
1′=(a
3+2a
2+2a
1+2b
1+b
2+4)/8 （2）
a
2′=(a
3+a
2+a
1+b
1+2)/4 (3)
a
3′=(2a
4+3a
3+a
2+a
1+b
1+4)/8 (4)
A in formula
1', a
2' and a
3' be a1, a2, the result of a3 after after filtering, a4 point is not processed; When b point value is carried out to filtering, only need in filter, change a point pixel of relevant position in formula (2), (3), (4) into b point pixel, b point pixel changes a point pixel into; Be the filtering mode that b1 is adopted to a1, b2 adopts the filtering mode of a2, and b3 adopts the filtering mode of a3, and b4 adopts the filtering mode of a4 to carry out filtering; Work as a, when the position of b is neighbouring, as shown in Cv in Fig. 3, only need to be by a, b is 90-degree rotation simultaneously, and Cv just becomes the form the same with C, and when then adjacent according to left and right, C situation is carried out filtering.
Execution mode is further in the present invention, it is all the situation of edge block for described adjacent block, the building form of continuing to use equally described C piece obtains C piece, if the pixel intensity in C centre position has obvious saltus step, this blocking effect of C is simulated with a two-dimentional step function blk, as shown in formula (5):
In formula, numerical value 1/2 and-1/2 has represented the alias in the middle of image block C; First the block effect intensity of the every a line centre pixel in C piece is assessed, as formula (6), used c (i, j) to represent the pixel value that i is capable, j is listed as in C here.
β=[c(i,6)-3×c(i,5)+3×c(i,4)-c(i,3)]/2（6）
In formula, β represents the block effect intensity in C piece centre position; Then select an applicable smooth function to replace the step function that produces blocking effect, the detailed information of considering edge block is more, and have the position texture structure complexity of blocking effect, therefore only boundary pixel is carried out slight smoothly, the smooth function of employing is as formula (7):
In formula, β level obtains according to block effect intensity β, can control adaptively the shape of smooth function, when β level value more hour smooth function approach original step function; Smooth function can change border pixel values in Min. ground, reduces the brightness saltus step on image block border, can not cause the fuzzy of image; In the time processing, this smooth function is carried out to discretization, construct one dimension smooth function, as shown in formula (8):
de_blk(j)=[f(-49)f(-35)f(-21)f(-7)f(7)f(21)f(35)f(49)]
c'(i,j)=c(i,j)+β×[de_blk(j)-blk(j)] （8）
Here, de_blk (j) is the array by 1 × 8 size of f (x) generation, c'(i, j) be the capable result after filtered of the i of C, here c (i, j) refer to the capable all elements of i of C, blk(j) refer in C that a line is modeled into the value after blk arbitrarily, each in this sampled images is passed through after level and smooth can effectively remove blocking effect.
Execution mode is further in the present invention, to the blocking effect between described edge block and flat block and between edge block and texture block, in order not make entire image too fuzzy because filtering causes, this two classes blocking effect is not processed.
Execution mode is further in the present invention, to the border between the blocking effect between described texture block and texture block and flat block, adopt the similar simplification loop block-eliminating effect filtering method of situation of the blocking effect between flat block, when difference is filtering, the pixel of a3 position and b3 position is not carried out to filtering processing.
Like this, after filtering after, finally just can obtain filtered deblocking effect image.
As shown in Figure 4, for the inventive method and existing method are to PQ=10, bpp=0.36 carries out filtering deblocking effect, the comparison of acquired results through jpeg compressed image.Compression quality parameter when wherein PQ value is compressed image for employing JPEG.
Fig. 5 is the result comparison that obtains PSNR value and the compressed image PSNR value of image after the inventive method deblocking effect.
The invention is not restricted to above-mentioned execution mode, those skilled in the art make to any apparent improvement of above-mentioned execution mode or change, can not exceed the protection range of design of the present invention and claims.
Claims (7)
1. the low bit compression image deblocking effect method based on visual sensitivity and airspace filter, is characterized in that, described method comprises the steps:
S1: the distorted image of input is divided into flat region, marginal zone and texture area according to visual sensitivity characteristic;
S2: select Sobel Operator to extract the gradient information of image slices vegetarian refreshments, entire image is being carried out to after Grad asks for, adopt three kinds of Component Models, setting threshold, is divided into marginal zone, texture area and flat site by image;
S3: respectively according to 8 × 8 big or small piecemeals, distorted image region is just further divided into: flat block, edge block and texture block by three regions of distorted image;
S4: again according to the neighbouring relations of the type of region unit and piece and piece, blocking effect type in distorted image is divided into: the blocking effect between the blocking effect between blocking effect, edge block between flat block, blocking effect, flat block and the edge block between texture block, the blocking effect between the blocking effect between edge block and texture block and texture block and flat block;
S5: for the blocking effect between flat block, adopt the loop block-eliminating effect filtering algorithm in video compression standard H.264, remove the blocking effect between flat block with the filter of filtering strength maximum, only selecting adjacent block in the time of filtering is all the border of flat block;
S6: for the blocking effect between edge block, adopt a kind of self-adaptive smooth filtering mode to carry out filtering, first described filtering assessed the block effect intensity of the every a line centre of block boundary pixel, then selects the smooth function of an applicable adaptive control by block effect intensity control to carry out filtering to blocking effect;
S7: the blocking effect between edge blocks and flat block and between edge block and texture block, in order not make entire image too fuzzy because filtering causes, does not process for this class blocking effect;
S8: for the border between the blocking effect between texture block and texture block and flat block, adopt the similar simplification loop block-eliminating effect filtering method of situation of the blocking effect between flat block.
2. the low bit compression image deblocking effect method based on visual sensitivity and airspace filter as claimed in claim 1, it is characterized in that, in described step S2, select Sobel Operator to extract the gradient information of image slices vegetarian refreshments, adopt four direction template, be respectively horizontal direction, vertical direction, the template of 45 degree directions and 135 degree directions, for piece image, first carry out convolution with this image respectively by this four direction template, obtain the Grad on image four direction, and select in four Grad maximum one, be denoted as gmax; Then adopt three kinds of Component Models that image is divided into fringe region, texture region and flat site; Concrete steps are:
(1), calculated threshold, TH1=γ × gmax and TH2=δ × gmax, TH1, TH2 are height threshold value, as the foundation of differentiating pixel Texture classification, γ span is between 0.1～0.2, δ span is between 0.01～0.1;
(2), adopt following formula that image pixel is divided into edge, texture or flat site;
Wherein, G (i, j) is the Grad of each pixel in image, and in the time that the Grad of pixel (i, j) position is greater than TH1, this pixel is just divided into marginal zone; In the time that the Grad of pixel (i, j) position is less than TH2, this pixel is just divided into flat region; If the Grad of pixel (i, j) position between TH1 and TH2 time this pixel be just divided into texture area.
3. the low bit compression image deblocking effect method based on visual sensitivity and airspace filter as claimed in claim 1, it is characterized in that, describedly by three regions of distorted image according to the concrete sorting technique of 8 × 8 big or small piecemeals be respectively: suppose a, b is respectively the flat block of two 8 × 8 pixel sizes, and a, the position relationship of b is that left and right is adjacent, the right side of a 4 is listed as with a left side 4 row pixels of b and forms an image block C, the blocking effect in the middle of a, b will be intactly retained in C, therefore to a, between b, the filtering of blocking effect is that C is carried out to filtering operation actually; Work as a, when the position relationship of b is neighbouring, form image block Cv; If a, b is flat block, and the blocking effect in C is the blocking effect between flat block, if a, b one of them be flat block another be edge block, the blocking effect in C is the blocking effect between flat block and edge block, can obtain by that analogy 6 kinds of blocking effect classifications.
4. the low bit compression image deblocking effect method based on visual sensitivity and airspace filter as claimed in claim 3, it is characterized in that, the method for described composition image block C and Cv, establishing piece C is the new piece of adjacent block a and b composition, if in C, a line 8 pixels are from left to right expressed as successively arbitrarily: a4, a3, a2, a1, b1, b2, b3, b4; In C, with the row indicating in black dotted lines, adopt formula to carry out below filtering to any a line in C so:
a
1′=(a
3+2a
2+2a
1+2b
1+b
2+4)/8
a
2′=(a
3+a
2+a
1+b
1+2)/4
a
3′=(2a
4+3a
3+a
2+a
1+b
1+4)/8
A in formula
1', a
2' and a
3' be a1, a2, the result of a3 after after filtering, a4 point is not processed; When b point value is carried out to filtering, only need in filter, change a point pixel of relevant position in formula into b point pixel, b point pixel changes a point pixel into; Be the filtering mode that b1 is adopted to a1, b2 adopts the filtering mode of a2, and b3 adopts the filtering mode of a3, and b4 adopts the filtering mode of a4 to carry out filtering; Working as a, when the position of b is neighbouring, is Cv, only need to be by a, and b is 90-degree rotation simultaneously, and Cv just becomes the form the same with C, and when then adjacent according to left and right, C situation is carried out filtering.
5. the low bit compression image deblocking effect method based on visual sensitivity and airspace filter as claimed in claim 4, it is characterized in that, it is all the situation of edge block for described adjacent block, the building form of continuing to use equally described C piece obtains C piece, if the pixel intensity in C centre position has obvious saltus step, this blocking effect of C is simulated with a two-dimentional step function blk, shown in following formula:
In formula, numerical value 1/2 and-1/2 has represented the alias in the middle of image block C; First the block effect intensity of the every a line centre pixel in C piece is assessed, following formula, uses c (i, j) to represent the pixel value that i is capable, j is listed as in C here.
β=[c(i,6)-3×c(i,5)+3×c(i,4)-c(i,3)]/2
In formula, β represents the block effect intensity in C piece centre position; Then select an applicable smooth function to replace the step function that produces blocking effect, the detailed information of considering edge block is more, and there is the position texture structure complexity of blocking effect, therefore only boundary pixel carried out slightly smoothly the following formula of smooth function of employing:
In formula, β level obtains according to block effect intensity β, can control adaptively the shape of smooth function, when β level value more hour smooth function approach original step function; Smooth function can change border pixel values in Min. ground, reduces the brightness saltus step on image block border, can not cause the fuzzy of image; In the time processing, this smooth function is carried out to discretization, construct one dimension smooth function, shown in following formula:
de_blk(j)=[f(-49)f(-35)f(-21)f(-7)f(7)f(21)f(35)f(49)]
c'(i,j)=c(i,j)+β×[de_blk(j)-blk(j)]
Here, de_blk (j) is the array by 1 × 8 size of f (x) generation, c'(i, j) be the capable result after filtered of the i of C, here c (i, j) refer to the capable all elements of i of C, blk(j) refer in C that a line is modeled into the value after blk arbitrarily, each in this sampled images is passed through after level and smooth can effectively remove blocking effect.
6. the low bit compression image deblocking effect method based on visual sensitivity and airspace filter as claimed in claim 5, it is characterized in that, to the blocking effect between described edge block and flat block and between edge block and texture block, in order not make entire image too fuzzy because filtering causes, this two classes blocking effect is not processed.
7. the low bit compression image deblocking effect method based on visual sensitivity and airspace filter as claimed in claim 5, it is characterized in that, to the border between the blocking effect between described texture block and texture block and flat block, adopt the similar simplification loop block-eliminating effect filtering method of situation of the blocking effect between flat block, when difference is filtering, the pixel of a3 position and b3 position is not carried out to filtering processing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410112243.4A CN103905822A (en) | 2014-03-25 | 2014-03-25 | Low-bit compressed image deblocking method based on visual sensitivity and spatial filtering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410112243.4A CN103905822A (en) | 2014-03-25 | 2014-03-25 | Low-bit compressed image deblocking method based on visual sensitivity and spatial filtering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103905822A true CN103905822A (en) | 2014-07-02 |
Family
ID=50996944
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410112243.4A Pending CN103905822A (en) | 2014-03-25 | 2014-03-25 | Low-bit compressed image deblocking method based on visual sensitivity and spatial filtering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103905822A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796624A (en) * | 2019-10-31 | 2020-02-14 | 北京金山云网络技术有限公司 | Image generation method and device and electronic equipment |
| CN114885165A (en) * | 2022-06-30 | 2022-08-09 | 深圳比特微电子科技有限公司 | Image frame loop filtering method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120177301A1 (en) * | 2011-01-10 | 2012-07-12 | Korea Advanced Institute Of Science & Technology | Method and apparatus for removing image blocking artifact by using transformation coefficient |
| CN103475876A (en) * | 2013-08-27 | 2013-12-25 | 北京工业大学 | Learning-based low-bit-rate compression image super-resolution reconstruction method |
-
2014
- 2014-03-25 CN CN201410112243.4A patent/CN103905822A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120177301A1 (en) * | 2011-01-10 | 2012-07-12 | Korea Advanced Institute Of Science & Technology | Method and apparatus for removing image blocking artifact by using transformation coefficient |
| CN103475876A (en) * | 2013-08-27 | 2013-12-25 | 北京工业大学 | Learning-based low-bit-rate compression image super-resolution reconstruction method |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796624A (en) * | 2019-10-31 | 2020-02-14 | 北京金山云网络技术有限公司 | Image generation method and device and electronic equipment |
| CN110796624B (en) * | 2019-10-31 | 2022-07-05 | 北京金山云网络技术有限公司 | Image generation method and device and electronic equipment |
| US11836898B2 (en) | 2019-10-31 | 2023-12-05 | Beijing Kingsoft Cloud Network Technology Co., Ltd. | Method and apparatus for generating image, and electronic device |
| CN114885165A (en) * | 2022-06-30 | 2022-08-09 | 深圳比特微电子科技有限公司 | Image frame loop filtering method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11381812B2 (en) | Boundary filtering method for intra prediction | |
| CN105608676B (en) | The Enhancement Method and device of a kind of video image | |
| Maggioni et al. | Video denoising, deblocking, and enhancement through separable 4-D nonlocal spatiotemporal transforms | |
| Singh et al. | Super-resolving noisy images | |
| WO2016206087A1 (en) | Low-illumination image processing method and device | |
| US9183617B2 (en) | Methods, devices, and computer readable mediums for processing a digital picture | |
| CN103475876B (en) | A kind of low bit rate compression image super-resolution rebuilding method based on study | |
| CN105046658A (en) | Low-illumination image processing method and device | |
| CN105096280A (en) | Method and device for processing image noise | |
| TWI477153B (en) | Techniques for identifying block artifacts | |
| JPH08186714A (en) | Noise removal of picture data and its device | |
| CN108564597B (en) | Video foreground object extraction method fusing Gaussian mixture model and H-S optical flow method | |
| CN101877123A (en) | Image enhancement method and device | |
| EP2816526B1 (en) | Filtering method and apparatus for recovering an anti-aliasing edge | |
| CN104469386A (en) | Stereoscopic video perception and coding method for just-noticeable error model based on DOF | |
| CN102611831B (en) | Method for reducing compressed image encoding noise | |
| Yan et al. | Method to Enhance Degraded Image in Dust Environment. | |
| CN103313047A (en) | Video coding method and apparatus | |
| CN105631854A (en) | FPGA platform-based self-adaptive image definition evaluation algorithm | |
| CN109345479B (en) | Real-time preprocessing method and storage medium for video monitoring data | |
| CN107341459A (en) | A kind of Face datection image pre-processing method based on fractional calculus | |
| CN103905822A (en) | Low-bit compressed image deblocking method based on visual sensitivity and spatial filtering | |
| JP2006148878A (en) | Method for classifying pixels in image | |
| KR101615479B1 (en) | Method and apparatus for processing super resolution image using adaptive pre/post-filtering | |
| CN110298803A (en) | A kind of compressed sensing based image band noise remove method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20140702 |
|
| RJ01 | Rejection of invention patent application after publication |