The human retina adaptively processes visual data even if the environmental or imaging conditions are far below ideal. The unique ability of eyes to adapt to poor light and environmental conditions for processing of low-level information is important for edge detection and recognition. This is accomplished through different responses of photoreceptors and adaptive center-surround responses of cells organized in four major layers. We present a structured neural network for image and edge enhancement that is derived from the model of biologcal understanding of retinal layers. The image is described by a set of interconnected neurons with their values equal to the gray-level values of corresponding pixels. The first-order and second-order contrast links are defined among the neurons, which are analyzed for changes in their values in the adaptive constrained environment. Each selected neuron is analyzed only once per iteration in which its value may be readjusted by incrementing or decrementing the current value. As a result, at the end of each iteration, the image data are reorganized for better contrast and image features. We present the structure and algorithm of the proposed neural network with various experimental results showing the capability of such a network to enhance the gray-level images. The method presented is suitable for computerized digital processing of gray-level images for enhancement.