Postprocessing of decoded color images by adaptive linear filtering
Yükleniyor...
Dosyalar
Tarih
2003-02
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Elsevier Science
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
This paper presents an image adaptive linear filtering method for the reconstruction of the RGB (red, blue, green) color coordinates of a pixel from the lossy compressed luminance/chrominance color coordinates. In the absence of quantization noise, the RGB coordinates of a pixel can be perfectly reconstructed by employing a standard, fixed filter whose support includes only the luminance/chrominance coordinates at the spatial location of the pixel. However, in the presence of quantization noise, a filter with a larger support, that also spatially extends over the luminance/chrominance coordinate planes, is capable of exploiting the statistical dependence among the luminance/chrominance coordinate planes, and thereby yields more accurate reconstruction than the standard, fixed filter. We propose the optimal (in the minimum mean squared error sense) determination of the coefficients of this adaptive linear filter at the image encoder by solving a system of regression equations. When transmitted as side information to the image decoder, the filter coefficients need not incur significant overhead if they are quantized and compressed intelligently. Our simulation results demonstrate that the distortion of the decompressed color coordinate planes can be reduced by several tenths of a dB with negligible overhead rate by the application of our image adaptive linear filtering method.
Açıklama
Anahtar Kelimeler
Color image coding, Color image processing, Adaptive linear filter, Wavelet, Adaptive filtering, Image coding, Image reconstruction, Vector quantization, Adaptive linear filters
Kaynak
Signal Processing: Image Communication
WoS Q Değeri
Q2
Scopus Q Değeri
Q2
Cilt
18
Sayı
2
Künye
Bayazit, U. (2003). Postprocessing of decoded color images by adaptive linear filtering. Signal Processing: Image Communication, 18(2), 91-101. doi:10.1016/S0923-5965(02)00083-8