Apparatus and Method for Image Resolution Conversion Using Patterns

Apparatus for Verifying Color Filter Array Performance Based on RRGB, RGGB, and RGBB Pattern Training Filters

This technology is an image resolution conversion device that designs training filters for reconstruction by separating and downsampling RRGB, RGGB, and RGBB pattern training images by channel and comparing them with actual pixel values. It then verifies array performance by using these filters to reconstruct captured images from candidate color filter arrays and comparing them to the originals.

While the Bayer RGGB pattern is primarily used in digital camera image sensors, each pixel captures only one color, requiring the remaining colors to be reconstructed via interpolation. Nearest-neighbor, bilinear, and bicubic interpolation are prone to causing aliasing in color planes, and there has been a lack of objective means to compare the performance of different color filter array patterns.

By applying the same training filter to various color filter array patterns and quantitatively comparing the reconstructed images with the originals, this technology enables efficient analysis of pattern-specific performance and provides verification results. It can be applied to simulation tools for image sensor design, camera module development, and sensor pattern selection, allowing for the optimal color filter array to be chosen before prototyping.

‍

‍

Key Features:

• A database storing RRGB, RGGB, and RGBB patterns of multiple training and target images, along with actual R, G, and B pixel values.
• A training filter design module that separates training image patterns by R, G, and B channels and downsamples them to form low-resolution images.
• A filter design unit that designs training filters by minimizing the squared error between the actual and reconstructed pixel values of the training images.
• A performance verification module that reconstructs target images from a selected color filter array image sensor using the training filters and compares them with the originals.

‍