Optical remote sensing data processing is proposed for the airborne images of high spectral and spatial resolution. Optimization techniques are undertaken to gain information about spatial distribution of the pixels on the hyperspectral images and the texture of the forest stands of different species and ages together with reducing redundancy of the spectral channels used. The category of neighborhood of pixels for particular forest classes and the step up method of selecting optimal spectral channels are employed in the relevant processing procedures. We present examples of pattern recognition for the forests as a result of separating pixels, which characterize the sunlit tops, shaded space and intermediate cases of the Sun illumination conditions on the hyperspectral images.