AN EFFICIENT IMAGE COMPRESSION ALGORITHM FOR BAYER-PATTERN IMAGES

Abstract

In this paper, an efficient image compression algorithm for Bayer-pattern images is introduced. In Bayer pattern images each pixel is represented by 8-bit rather than 24-bit for colour images. The proposed algorithm separates the colour information from the Bayer-pattern image data before performing the compression operation. It partitions each colour plane into fixed size blocks. Each block is classified as a low or high intensity variation block using the standard deviation which compared against a predefined threshold. The low intensity variation blocks are encoded by its average while, the other blocks are encoded using the minimum, quantisation step and the bit plane. The quantisers that used in this paper are from 32 and 16 levels types. The effect of the block size on the output bit rate and image accuracy is studied. The results clarify that a good performance is achieved using 4×4-pixel blocks. Using 32 and 16 quantisation levels a bit rates of 1.8 to 5.2 bpp with a PSNR of 30 to 45 dB can be achieved for a range of thresholds 10 to 0.5.

Keywords

• Image compression
• Bayer pattern
• Encoding
• Quantising