The data generated in wireless capsule endoscopy during the examination of the gastrointestinal tract is huge and demands large storage and processing time. One of the essential components of the capsule is the compression module that needs to be low power and a low complexity one. This paper aims to bring significant improvement in pre-processing, chroma subsampling and encoding stages of image compression. Among various chrominance based color spaces, YEF being a best alternative for the RGB endoscopic image is analyzed in terms of statistical measures. New chroma subsampling patterns are tested, which has shown improved data reduction. In the encoder part, the combination of proposed predictive coder and the modified Golomb Rice coder generated better compressed bit stream. The cubic spline interpolation is the method used for image reconstruction during decompression. The endoscopic images from databases namely Gastrolab and WEO Clinical Endoscopy Atlas are utilized in the experimentation. The results show that the proposed near-lossless compression scheme performs better than the competing techniques in terms of compression rate, peak signal to noise ratio and structural similarity index.