Integrating a Smooth Psychovisual Threshold into an Adaptive JPEG Image Compression

Abstract

The visual quality image output of JPEG compression is determined by quantization process. The popular quality factor in the extended JPEG image compression has been widely used to scale up the quantization tables. The scaling quantization table using quality factor is used to determine the quality image output. The scaling up on the quantization tables increases their values uniformly thus produces higher compression performance. However, the effects of the scaling up on the human visual system have not been taken into consideration. This paper examines the quantization table design based on adaptive psychovisual threshold and numerical analysis of the compression performance in terms of quality image reconstruction and average bit length of Huffman code. The comparison between extended JPEG image compression using the typical quality factor and quality scale of psychovisual threshold has been done. The experimental results of adaptive quantization tables based on psychovisual threshold show an improvement on the quality of image reconstruction at the lower average bit length of Huffman's code.

Psychovisual threshold on large Tchebichef moment for image compression

“…In this quantitative experiment, 80 images (24-bit RGB with 512×512 pixels) [13] are chosen to be tested and analyzed in order to develop psychovisual threshold in image compression. In past project work, psychovisual threshold has been investigated on 8×8 discrete transform [14][15][16]. First, the sample RGB image is converted into YCbCr colour space.…”

“…A large 256×256 quantization table consists of frequency order zero to the maximum of frequency order 510. 22 23 23 23 22 21 20 17 16 13 11 8 5 4 2 2 22 22 22 20 20 22 22 23 23 23 23 21 20 17 16 13 11 8 6 4 2 2 22 22 20 19 22 22 23 23 23 23 21 20 17 16 13 11 8 6 3 2 2 22 20 19 22 22 23 23 23 23 21 20 17 16 12 11 7 6 3 2 20 19 22 22 23 23 23 23 21 20 17 16 12 12 7 6 3 2 20 22 22 23 23 23 23 20 20 17 16 12 12 7 7 3 2 There are 2 16 = 65536 quantization values from a 256×256 quantization table. The order 0 resides in the top left most corner of the quantization matrix order of Q(0,0).…”

Event-Driven Extraction of HTTP Payload From Concurrent Streams

“…In some cases compression is done before the object get transmitted. If chunked encoding is applied on compressed blocks [3], the sending process can be carried out simultaneously with the compression algorithm, thus the client may receive the first blocks before the whole compression process is finished.…”