Öz
We present in this work, an improved JPEG-based image compression method. The proposed
method appeals a so-called QRLE (Quantized Run Length Encoding) technique where the couple of values
(‘zero’ followed by its run number) is replaced just by one value. In this work, we substitute the Huffman
encoding process by the QRLE method. The key idea of the QRLE method is to guarantee that the MSB bit of
the binary representation of the non-null values is ‘0’ while that of the ‘zero’ run number is ‘1’; this is done by
adding the value 2N-1 to that number where N is its binary representation length. However, given that the zero
run numbers could exceed (27-1≡127) value, provided that the RGB colored images are 8 bits long, the MSB is,
necessarily, ‘1’. Even worse, the zero run number can be greater than 255 which implies occupying more than
one octet. To solve that; the basic idea is to treat the zero run numbers of the AC zig-zag coefficients, according
to 3 pre-defined ranges as follows: [1..127], [128..255], and [256..maximum. For the first range, ‘128’ value is
added to the zero run numbers; while for the second one, the zero run numbers are unchanged but preceded by
‘0’ value. Finally, for the third range, the zero run numbers are kept unchanged while preceded by the number of
octets of their binary representation. In terms of obtained results, comparing the proposed method with the
classical JPEG, the reconstitution values remain the same for both technics. On the other hands, the compression
ratio is considerably improved rising from 1:16.35 to 1:22.62 being provided an RGB image with 3 times a
typical quantization matrix (for a quality of 50% as specified in the original JPEG Standard) while the PSNR is
around 26.215both algorithms.
Anahtar Kelimeler
Ayrıntılar
| Birincil Dil | İngilizce |
|---|---|
| Konular | Mühendislik |
| Bölüm | Makaleler |
| Yazarlar | |
| Yayımlanma Tarihi | 31 Aralık 2022 |
| Yayımlandığı Sayı | Yıl 2022Cilt: 21 |
