CN103903285B - Thumbnail extraction method of PNG image and BMP image - Google Patents

Abstract

A method for extracting thumbnails of PNG images and BMP images. The present application discloses a method for extracting a thumbnail image of a PNG image, comprising: for a non-interlaced PNG image, opening up a buffer of the size of the space occupied by two rows of pixels for storing the pixel values of the current row and the previous row before decoding; The PNG image is decoded row by row starting from the first row, and downsamples the decoded reconstructed pixel values of each row to obtain the corresponding row pixel values of the target thumbnail; wherein, when the row-by-row decoding is performed, After the decoding of the current line except the first line is completed, the pixel values of the previous line before decoding are moved out of the buffer, and then the pixel values of the next line before decoding are written into the buffer, and the next line is used as the current line line to decode. By applying this application, the memory overhead of the system can be saved.

Description

A Method for Extracting Thumbnails of PNG Images and BMP Images

technical field

The present application relates to image processing technology, in particular to a method for extracting thumbnails of BMP images.

Background technique

A personal computer can access multimedia files (video, music, pictures, etc.) stored on its own hard disk, other external devices, and the network, and provide users with a thumbnail preview function for browsing the overview of these files. Display devices based on embedded systems such as digital TVs and mobile phones also have the same preview function. Wherein, the preview picture can provide the user with an intuitive overview of the content of the picture, and has high practical value.

Since the image files are small, the storage medium can store a large number of image files. This requires that the display device can quickly process the original image to generate thumbnail information, and quickly display the thumbnail.

At present, for obtaining thumbnails of image files, existing methods include hardware implementations and software implementations. Among them, when implemented by hardware, although its speed is fast, its scalability, fault tolerance and compatibility with multiple formats are poor. When using software to extract thumbnails, different thumbnail extraction methods will be adopted according to different image source formats.

Specifically, for thumbnails in PNG (Portable Network Graphics Format) format, existing methods mainly decode the original image directly, and then scale the original image to obtain thumbnail information. For thumbnails in BMP (Bitmap) format, the existing method is mainly to directly decode the original image, and then scale the original image to obtain the thumbnail information.

For the thumbnail extraction in PNG format performed by the above-mentioned software method, the original image is usually directly decoded and then scaled. For larger images, full image decoding will take up more system memory. For example, for a 1920x1080 PNG image whose color space is RGBA, decoding the whole image and storing the decoded image will consume 8M of system memory, which is not conducive to extracting thumbnails of multiple images.

For the thumbnail extraction in BMP format performed by the above-mentioned software method, the original image is usually directly decoded and then scaled. Since images in BMP format are often relatively large (such as 24bit BMP images are the most common), decoding will take up a lot of memory and take a long time, which is difficult to meet the real-time display requirements. For example: for a 24bit BMP image with a resolution of 1920x1080, decoding and scaling the entire image will consume 6M of system memory, and it is very time-consuming to read and store each pixel, which cannot meet the requirements of real-time display and is not conducive to Extract thumbnails for multiple images.

Contents of the invention

The application provides a method for extracting a thumbnail of a PNG image and a method for extracting a thumbnail of a BMP image, which can greatly save system memory consumption and increase the speed of extracting thumbnails.

A method for extracting thumbnails of PNG images, comprising:

For non-interlaced PNG images, a buffer of the space occupied by 2 rows of pixels is opened to store the pixel values before decoding of the current row and the previous row;

Decoding the PNG image row by row starting from the first row, and downsampling the decoded reconstructed pixel values of each row to obtain the corresponding row pixel values of the target thumbnail; wherein, when the row-by-row decoding is performed , after the decoding of the current line except the first line is completed, the pixel value of the previous line before decoding is moved out of the buffer, and then the pixel value of the next line before decoding is written into the buffer, and the next line is used as The current line to decode.

Preferably, the method further includes: after down-sampling the reconstructed pixel values of the current row, the method further includes: determining whether the decoded reconstructed pixel values need color mixing processing according to the input parameters of the image display device, and if necessary , the downsampled pixel values of the current line are mixed and then input to the display device; otherwise, the downsampled pixel values of the current line are directly input to the display device.

Preferably, when the input data format required by the display device is R, G, B, A four-channel pixel values, it is determined that color mixing processing is not required.

Preferably, when the display device requires the input data format to be R, G, and B three-channel pixel values, it is determined that color mixing processing is required;

The color mixing process is: calculate according to the R, G, B, A pixel values of each reconstructed pixel Wherein, BG is the background color, black background BG=0, white background BG=1, R′, G′, B′ are pixel values after color mixing, and are used for the down-sampling process.

Preferably, when the display device requires the input data format to be Y, Cb, and Cr three-channel pixel values, it is determined that color mixing processing is required;

The color mixing process is: calculate according to the R, G, B, A pixel values of each reconstructed pixel And calculate according to R', G', B' Wherein, BG is the background color, black background BG=0, white background BG=1, Y, Cb, Cr are pixel values after color mixing, and are used for the down-sampling process.

Preferably, the calculation For: calculate

Preferably, the calculation based on R', G', B' To: Correspond to all the values of R', G', B' in advance, calculate the corresponding Y, Cb, Cr, and save the calculation results corresponding to the corresponding values of R', G', B', in the calculation , select Y, Cb, and Cr corresponding to the current R', G', and B' values as the calculation results.

Preferably, the method further includes: for the interleaved PNG image, decoding an image whose size is closest to the target thumbnail, and downsampling the decoded pixel values to obtain the target thumbnail.

A method for extracting a thumbnail of a BMP image, comprising: for a BMP image in a non-RLE compressed format, any pixel of the corresponding target thumbnail, calculating the byte position of the corresponding BMP image data;

Read the color table index at the calculated byte position and determine the corresponding pixel value according to the color table index, or directly read the RGB data at the byte position and write it into the target thumbnail The position corresponding to any pixel.

Preferably, the byte position of the BMP image data corresponding to the calculation includes:

byte=T _start +linesize×((i-1)×H _BMP )/H _THU +((j-1)×W _BMP )×pixelsize/(W _THU ×8)

Wherein, T _start is the data start position in the basic information of the BMP image, linesize is the number of bytes of each row of pixels in the basic data of the BMP image, and pixelsize is the bit of each pixel in the basic data of the BMP image number.

It can be seen from the above-mentioned technical scheme that in the present application, for non-interlaced PNG images, a buffer of the size of the space occupied by 2 rows of pixels is opened up to store the pixel values before decoding of the current row and its previous row; Start the decoding process row by row, and down-sample the decoded reconstructed pixel values of each row to obtain the corresponding row pixel values of the target thumbnail image. When decoding row by row, after the decoding of the current row except the first row is completed, the pixel value before decoding of the previous row is moved out of the buffer, and then the pixel value of the next row before decoding is written into the buffer, and the next row Decode as the current line. In this way, during the whole decoding process, only the pre-decoding pixel values of 2 lines of images need to be stored, which greatly saves the system memory consumption during decoding.

Description of drawings

Fig. 1 is the schematic flow chart of the thumbnail extraction method of PNG image in the embodiment of the present application;

FIG. 2 is a schematic diagram of a method for extracting a thumbnail of a BMP image in the present application.

detailed description

In order to make the purpose, technical means and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings.

FIG. 1 is a schematic flowchart of a method for extracting a thumbnail of a PNG image in an embodiment of the present application. As shown in Figure 1, the method includes:

Step 101, judge whether current image is the image of PNG format, if yes, then continue to execute step 102, if not then exit;

Step 102, parse the header information of the PNG image to obtain the basic information of the image.

Among them, the acquired basic information includes: width W _ORG and height H _ORG of the original image, bit depth, color type and interleaving type, etc.

Step 103, if the determined file interleaving type is an interlaced PNG image, execute step 104; if it is a non-interleaved PNG image, execute steps 105-106.

Step 104, for the interlaced PNG image, decode an image in the PNG image file whose size is closest to the target thumbnail, and down-sample the image to obtain the target thumbnail.

The processing of this step is the same as the existing method. Assuming that I _BASE represents the pixel value of an image closest to the size of the target thumbnail in the PNG image file, and the width and height dimensions of the target thumbnail obtained by downsampling are W _REL and H _REL respectively, then the pixel value of each pixel of the target thumbnail Value I _REL (i, j) = I _BASE (HIndex[i], WIndex[j]), where HIndex[] and WIndex[] are row index array and column index array respectively, and HIndex[] indicates the row of I _REL The mapping relationship between the index and the row index of I _BASE , WIndex[] indicates the mapping relationship between the column index of I _REL and the column index of I _BASE , these mapping relationships are calculated in advance according to the size of the target thumbnail and the size of the PNG image, Specifically, it can be implemented according to an existing method, which will not be repeated here.

Step 105: For the non-interlaced PNG image, open up a buffer with the space required for two lines of pixels, which is denoted as a prediction buffer, and the stored pixel values of the upper and lower adjacent lines are respectively denoted as line0 and line1.

Step 106: Decode the PNG image line by line starting from the first line, and down-sample the decoded reconstructed pixel values of each line to obtain the corresponding line pixel values of the target thumbnail image.

When performing decoding processing line by line, it can be performed in the following manner:

A. Write the pre-decoded pixels of the first two lines into the prediction buffer, decode the first line, and use the second line as the current line;

B. Decoding the current row to obtain decoded reconstructed pixels;

C. Move the pixel value before decoding of the previous line of the current line out of the prediction buffer;

D. Write the pre-decoding pixel values of the next line of the current line into the prediction buffer, and return the next line as the current line to step B until the decoding of the entire image is completed.

In the above processing, when the previous line of the current line is moved out of the prediction buffer and the next line is written into the prediction buffer, preferably, after the previous line is moved out of the prediction buffer, the pixel values of the current line before decoding can be moved to the previous line and moved out At the position before the prediction buffer, write the undecoded pixels of the next line into the position before the current line moves. In this way, the data that is moved out of the prediction buffer each time is the data at the same location. Here, the reason why two lines of pixels are stored in the prediction buffer is to consider that when decoding the current line, the pre-decoding pixel value of the previous line may be used as a prediction reference.

After each decoding process of the current row is completed in step B, the decoded and reconstructed pixel values of the current row may be down-sampled to obtain the corresponding row pixel values of the target thumbnail image. That is to say, the down-sampling process can be performed synchronously with the execution of steps C, D and even the next step B, so as to increase the thumbnail acquisition speed.

Wherein, the operation of performing down-sampling is the same as the down-sampling in the aforementioned step 104 . Assuming that I _BASE represents the decoded and reconstructed pixel value of the current row, and the width and height dimensions of the target thumbnail obtained by downsampling are W _REL and H _REL respectively, then the pixel value I _REL (i, j)=I of each pixel of the target thumbnail _BASE (HIndex[i], WIndex[j]), where HIndex[] and WIndex[] are the row index array and column index array respectively, and HIndex[] indicates the mapping relationship between the row index of I _REL and the row index of I _BASE , WIndex[] indicates the mapping relationship between the column index of I _REL and the column index of I _BASE . These mapping relationships are calculated in advance according to the size of the target thumbnail and the size of the PNG image. Specifically, it can be implemented according to the existing method, here is No longer.

When downsampling the reconstructed pixels of the current row, only the pixel values corresponding to the target thumbnail are copied to the target thumbnail, and the rest of the pixel values are not processed. When the decoding of the pixel values of all rows is completed and the downsampling process is implemented, the loop steps B-D are stopped.

So far, the method for extracting a thumbnail of a JPEG image in this application ends.

In addition, considering the different requirements of the input parameters of different display devices, it may be necessary to perform color mixing on the reconstructed and downsampled pixel values (that is, to mix the transparent color channel into the RGB three channels), otherwise the image may appear aliasing effect.

Based on this, preferably, step E can be further performed after step B is executed each time: according to the input parameters of the image display device, it is determined whether the decoded RGBA data needs color mixing processing, and if color mixing processing is required, the down-sampled Perform corresponding color mixing processing on the pixel value, and input the pixel value after color mixing processing to the display device; otherwise, directly input the downsampled pixel value to the display device.

Specifically, this application provides the following two specific methods for judging whether to perform color mixing processing and the corresponding color mixing processing:

1) If the display device requires the input of four-channel pixel values of R, G, B, and A, it is determined that color mixing processing is not required, and the decoded R, G, B, and A are directly input to the display device;

2) If the display device requires the input of R, G, and B three-channel pixel values, it is determined that color mixing processing is required. The specific color mixing processing is: perform color mixing processing on the R, G, and B values of each pixel value, and the mixed color The pixel values are R', G', B' respectively:

Color mixing formula:

Because the above formula has division, preferably, it can be further optimized by shifting:

R'=[(R×A)>>8]+(255-A)×BG

G'=[(G×A)>>8]+(255-A)×BG

B'=[(B×A)>>8]+(255-A)×BG

(where BG is the background color, black background BG=0, white background BG=1)

After testing, it is found that: using shift instead of division, the optimized thumbnail extraction time is 10% to 20% shorter than the thumbnail extraction time, and the loss of precision per pixel does not exceed 1, which has a significant effect on quickly extracting PNG image thumbnails Effect;

3) If the display device requires the input of three-channel pixel values of Y, Cb, and Cr, first perform color mixing processing on each pixel according to the color mixing formula in (1) above to obtain R', G', and B', and then The obtained R', G', and B' are converted into Y, Cb, and Cr values according to the following formula, and the Y, Cb, and Cr values are input into the display device as pixel values after color mixing.

At the same time, since the above formula has floating-point number multiplication, and the value range of R', G', B' is 0 to 255, so in order to further improve the calculation efficiency, it is preferable to pre-correspond to R', G', B' For all the values, calculate the corresponding Y, Cb, Cr, and save the calculation results corresponding to the corresponding R', G', B' values. R', G', and B' are selected, and the corresponding Y, Cb, and Cr values are determined by looking up the table. After testing and optimization, the thumbnail extraction speed can be increased by 50%.

It can be seen from the specific implementation of the above-mentioned application that in this application, the use of progressive decoding greatly reduces the memory overhead in the process of decoding PNG images, and can greatly reduce the amount of memory used; after testing, for a 1080p non-interlaced PNG image, this The memory usage of the application method is 2‰ of full-image decoding; for a 1080i interlaced PNG image, the memory usage of the application method is 50% of the existing method, and the decoding time is shortened to less than 60% of full-image decoding. The present invention combines the decoding and reconstruction process with the thumbnail image pixel extraction process to further reduce the computational complexity and time consumption.

FIG. 2 is a schematic diagram of a method for extracting a thumbnail of a BMP image in the present application. As shown in Figure 2, the method includes:

Step 201, judge whether current image is the image of BMP format, if yes, then continue to execute step 202, if not then exit;

Step 202, analyzing the header information of the BMP image to obtain the basic information of the image.

The basic information of the BMP image includes: image width W _BMP , height H _BMP , compression type, number of bits per pixel pixelsize, number of bytes of pixels per row linesize, color table, and data start position T _start .

Step 203, if the BMP image is an RLE compressed format, then execute step 206 to decode the image; otherwise execute step 204;

Step 204, calculating the mapping relationship between thumbnail index and BMP data.

Assume that the width of the target thumbnail is W _THU and the height is H _THU . Here, for the thumbnail whose row index is i (i=1,...,H _THU ) and column index is j (j=1,...,W _THU ) Thumbnail index, its corresponding BMP data byte position is calculated as follows:

byte=T _start +linesize×((i-1)×H _BMP )/H _THU +((j-1)×W _BMP )×pixelsize/(W _THU ×8)

In this way, only the following two items need to be calculated:

H _i =linesize×((i-1)×H _BMP )/H _THU

W _j =((j-1)×W _BMP )×pixelsize/(W _THU ×8)

The mapping relationship can be expressed as:

byte=T _sart +H _i +W _j

Step 205, for the thumbnail image I, for a certain pixel I _i,j , first obtain its corresponding data starting position byte by byte=T _start +H _i +W _j , and read the corresponding color table index at this position , and then look for the corresponding pixel value in the color table, or directly read the RGB data (24bit, 32bit BMP image) and write it to the corresponding pixel position of the target thumbnail.

Step 206, for the BMP image in the RLE compression format, first decode the entire image, and then perform down-sampling according to the scaling ratio, and obtain the final thumbnail image.

Wherein, the down-sampling process is the same as the down-sampling in the aforementioned PNG image processing, and will not be repeated here.

In the BMP image thumbnail extraction method of the above-mentioned application, the memory usage can be greatly reduced for common BMP format images, and the processing speed is faster; after testing: for a 1080p BMP image, the extraction resolution is 160x120 For the thumbnail image, the memory usage of the method of this application is only 1% of that of the full image decoding; the decoding time is shortened to less than 10% of that of the full image decoding.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the present invention. within the scope of protection.

Claims (7)

1. a method for extracting thumbnail images of PNG images, characterized in that, comprising: For non-interlaced PNG images, a buffer of the space occupied by 2 rows of pixels is opened to store the pixel values before decoding of the current row and the previous row; Decoding the PNG image row by row starting from the first row, and downsampling the decoded reconstructed pixel values of each row to obtain the corresponding row pixel values of the target thumbnail; wherein, when the row-by-row decoding is performed , after the decoding of the current line except the first line is completed, the pixel value of the previous line before decoding is moved out of the buffer, and then the pixel value of the next line before decoding is written into the buffer, and the next line is used as The current line is decoded; Wherein, after downsampling the reconstructed pixel values of the current row, the method further includes: according to the input parameters of the image display device, determining whether the decoded reconstructed pixel values need color mixing processing, and if so, downsampling the current row After the pixel value is processed by color mixing, it is input to the display device; otherwise, the pixel value after the downsampling of the current line is directly input to the display device. 2 . The method according to claim 1 , wherein, when the input data format required by the display device is R, G, B, and A four-channel pixel values, it is determined that color mixing processing is not required. 3. The method according to claim 1, wherein, when the data format required by the display device to be input is R, G, and B three-channel pixel values, it is determined that color mixing processing is required; The color mixing process is: calculate according to the R, G, B, A pixel values of each reconstructed pixel Wherein, BG is the background color, black background BG=0, white background BG=1, R', G', B' are the pixel values after color mixing, which are used for the down-sampling process. 4. The method according to claim 1, characterized in that, when the data format required by the display device to be input is Y, Cb, and Cr three-channel pixel values, it is determined that color mixing processing is required; The color mixing process is: calculate according to the R, G, B, A pixel values of each reconstructed pixel And calculated according to R', G', B' Wherein, BG is the background color, black background BG=0, white background BG=1, Y, Cb, Cr are pixel values after color mixing, and are used for the down-sampling process. 5. The method according to claim 3 or 4, wherein said calculating The method is: calculate 6. The method according to claim 4, characterized in that, the calculation according to R', G', B' To: Correspond to all the values of R', G', B' in advance, calculate the corresponding Y, Cb, Cr, and save the calculation results corresponding to the corresponding values of R', G', B', in the calculation , select Y, Cb, and Cr corresponding to the current R', G', and B' values as the calculation results. 7. The method according to claim 1, further comprising: for the interleaved PNG image, decoding an image closest to the size of the target thumbnail, and downsampling the decoded pixel value, Get the target thumbnail.