Iterative Optimization of Quarter Sampling Masks for Non-Regular Sampling Sensors

Abstract

Non-regular sampling can reduce aliasing at the expense of noise. Recently, it has been shown that non-regular sampling can be carried out using a conventional regular imaging sensor when the surface of its individual pixels is partially covered. This technique is called quarter sampling (also 1/4 sampling), since only one quarter of each pixel is sensitive to light. For this purpose, the choice of a proper sampling mask is crucial to achieve a high reconstruction quality. In the scope of this work, we present an iterative algorithm to improve an arbitrary quarter sampling mask which results in a continuous increase of the reconstruction quality. In terms of the reconstruction algorithms, we test two simple algorithms, namely, linear interpolation and nearest neighbor interpolation, as well as two more sophisticated algorithms, namely, steering kernel regression and frequency selective extrapolation. Besides PSNR gains of +0.31 dB to +0.68 dB relative to a random quarter sampling mask resulting from our optimized mask, visually noticeable enhancements are perceptible.

A Novel End-To-End Network for Reconstruction of Non-Regularly Sampled Image Data Using Locally Fully Connected Layers

“…This is achieved by using a sampling mask that repeats every 8×8 pixels. More precisely, we use an optimized mask of size 8×8 pixels from [6]. As a side effect, using periodically repeating masks can also be considered advantageous for the hardware manufacturing.…”

“…For VDSR-QS the same set of hyper-parameters as in [15] is used. Regarding the quarter sampling mask, we use the optimized quarter sampling mask from [6] because it shows an improved reconstruction quality for fij . This mask is of size 32×32 pixels and is repeated periodically until the respective reference image is covered.…”

“…For a proper reconstruction, high-quality reconstruction algorithms such as the frequency selective reconstruction (FSR) [5] need to be used in combination with optimized quarter sampling patterns such as those in [6]. FSR has shown to be a successful reconstruction scheme for various inpainting and extrapolation tasks [7]- [9] and showed best results for non-regular sampling and quarter sampling in [1], [5], [6]. An advancement of quarter sampling is called three-quarter sampling [10].…”

Novel Consistency Check For Fast Recursive Reconstruction Of Non-Regularly Sampled Video Data

“…Due to the non-regularity, visually disturbing aliasing artifacts that conventionally occur for regular sampling can be reduced [2,3,4]. For the reconstruction, frequency selective reconstruction (FSR) has shown to be a successful reconstruction scheme for various inpainting and extrapolation tasks [5,6] and gave best results for non-regular sampling and quarter sampling in [1,7,8]. Quarter sampling, as well as any nonregular sub-sampling, can be seen as a special case of compressed sensing [9,10] as has been shown in [7,11].…”

Enhanced Image Reconstruction From Quarter Sampling Measurements Using An Adapted Very Deep Super Resolution Network

“…The missing pixels need to be reconstructed from the sampled data. For a proper reconstruction, high-quality reconstruction algorithms such as the frequency selective reconstruction (FSR) [5] need to be used in combination with optimized quarter sampling patterns such as those in [6]. FSR has shown to be a successful reconstruction scheme for various inpainting and extrapolation tasks [7,8,9] and showed best results for nonregular sampling and quarter sampling in [1,5,6].…”