Image analysis for image compression suitability and real-time selection

• US 20040081357A1
• Filed: 10/29/2002
• Published: 04/29/2004
• Est. Priority Date: 10/29/2002
• Status: Active Grant

• Current Assignee: ATI Technologies ULC (Advanced Micro Devices, Inc.)
• Original Assignee: ATI Technologies ULC (Advanced Micro Devices, Inc.)
• Inventors: Oldcorn, David, Pomianowski, Andrew, Koduri, Raja
• Granted Patent: US 7,903,892 B2
• Time in Patent Office: Days
• US Class Current: 382/166
• CPC Class Codes: H04N 19/12 Selection from among a plurality of transforms or standards, e.g. selection between discrete cosine transform [DCT] and sub-band transform or selection between H.263 and H.264
• CPC Class Codes: H04N 19/154 Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion use of rate-distortion criteria H04N19/147
• CPC Class Codes: H04N 19/176 the region being a block, e.g. a macroblock
• CPC Class Codes: H04N 19/60 using transform coding

Abstract

The present invention provides an analysis method that uses error metrics to determine whether an image is suitable for compression. One embodiment of the present invention can make the determination in real-time. In one embodiment, analysis methods based on four error metrics are used to determine whether a compressed image should be used. The first metric is a signal-to-noise (SNR) error metric that prevents compressed images with low SNR from being used to represent original images. Another metric is detecting the geometric correlation of pixels within individual blocks of the compressed images. The third metric is used to determine whether color mapping in the compressed images are well-mapped. A final metric is a size metric that filters images smaller than a certain size and prevents them from being compressed. As most analysis methods are integral parts of the compression process, the present invention incurs little cost collecting error metric data.

Claims

1. A method of determining image compression suitability comprising the steps of:

performing a compression process on an image to generate a compressed image of said image whereby said image is divided into blocks;

gathering data from said image;

performing an error metric analysis with said data;

checking each of said blocks in said image; and

deciding whether to accept said compressed image as a suitable representation of said image.

2. The method of claim 1 wherein said step of performing an error metric analysis uses geometric correlation as said error metric.

3. The method of claim 2 wherein said step of performing a geometric correlation error metric analysis further comprises:

identifying all pixel values in each of said blocks;

performing a first check of said pixel values to ensure that pixel values progress in decreasing monotone fashion from left to right in each row,

performing a second check of said pixel values to ensure that pixel values progress in decreasing monotone fashion from top to bottom in each column; and

declaring a geometric correlation when said first and second checks are satisfied.

4. The method of claim 3 further comprising:

rotating the image by 90 degrees;

performing said first and second checks;

rotating the image by 180 degrees;

performing said first and second checks;

rotating the image by 270 degrees;

performing said first and second checks; and

declaring a geometric correlation when said all of said checks are satisfied.

5. The method of claim 1 wherein said step of performing an error metric analysis uses axis mapping as said error metric.

6. The method of claim 5 wherein said step of performing an axis mapping error metric analysis further comprises:

calculating axis mapping error between a source point and a mapped point on the color axis in the three-dimensional color space;

performing said steps of calculating for axis mapping errors for all source points in each of said blocks; and

summing all of said axis mapping errors using a summed square error method to produce a summed square error, and

determining if said summed square error is greater than a preset limit.

7. The method of claim 6 wherein said step of summing uses a root mean square method.

8. The method of claim 1 further comprising the steps of:

dividing an image into blocks;

gathering data from said image;

performing an error metric analysis with said data;

checking each of said blocks in said image; and

deciding whether said image is a suitable candidate for compression.

9. The method of claim 8 wherein said step of performing an error metric analysis uses signal-to-noise ratio as said error metric.

10. The method of claim 9 wherein said step of performing a signal-to-noise ratio error metric analysis further comprises:

decomposing said image into three color components;

transforming said blocks of said image into frequency domain for each of said three color components using the DCT method;

calculating values of low frequency blocks of said image; and

summing said values of low frequency blocks.

11. The method of claim 9 wherein said step of performing a signal-to-noise ratio error metric analysis further comprises:

checking whether said image is using the full range of color bit-depth.

12. The method of claim 8 wherein said step of performing an error metric analysis uses image size as said error metric.

13. The method of claim 12 where said image size is 64×

64 pixels.

14. The method of claim 1 wherein all of said steps are performed in real-time.

15. A method of determining image compression suitability comprising the steps of:

performing signal to noise ratio error metric analysis on an image;

applying DXTC compression to said image to produce a compressed image if said image passes said signal to ratio error metric;

performing color axis mapping error metric analysis on said image;

performing geometric correlation error metric analysis on said image;

performing size error metric analysis on said image; and

using said compressed image if said image passes said error metric analysis and geometric correlation error metric analysis and size error metric analysis.