Adaptive Edge-Preserving Filter for CMOS Image Sensor Cross-Talk Compensation | IEEE Conference Publication | IEEE Xplore

Adaptive Edge-Preserving Filter for CMOS Image Sensor Cross-Talk Compensation


Abstract:

This paper presents an adaptive edge-preserving filter (AEPF) to remove the pixel cross-talk phenomenon of CMOS image sensor. As is widely known, the pixel crosstalk is d...Show More

Abstract:

This paper presents an adaptive edge-preserving filter (AEPF) to remove the pixel cross-talk phenomenon of CMOS image sensor. As is widely known, the pixel crosstalk is due to the optical, electrical and architectural interferences and leads to the blocky formation after color interpolation. Although the conventional algorithms can eliminate the blocky formation, the edge false colors are also induced to degrade the image quality accordingly. Therefore, the AEPF is proposed to eliminate the blocky formation and to prevent the generation of edge false colors. Experimental results revealed that the AEPF can effectively remove the pixel cross-talk effect and result in good image quality without blocky formation and edge false colors
Date of Conference: 24-27 September 2006
Date Added to IEEE Xplore: 26 December 2006
CD:1-4244-0535-1
Conference Location: Teton National Park, WY, USA

1. Introduction

Recently, digital camera inside the mobile devices, such as cellular phone and PDA, has become a necessary feature to consumer electronic industrial. Those applications require low power consumption, pocket dimension, low cost, and easy integration. Therefore, CMOS (Complementary Metal Oxide Semiconductor) image sensor is more suitable than CCD (Charge Coupled Device) for the mobile devices to transfer the natural scenes into electrical signals. CMOS image sensor, however, is inherent with many kinds of noises. For instance, the reset noise and integration noise are related to the pixel-controlled circuit [3]. The dark current is induced by the electrical leakage. In addition, pixel crosstalk is generated from optical, electrical and architectural interferences. Moreover, the bad pixel and wounded pixel are resulted from material and manufacturing defects. In general, the bad/wounded pixel can be removed as an impulsive noise and the integration noise can be considered as a uniform noise. The dark current can be cancelled by measuring the current of optical black. The reset noise can be compensated through the correlated-double-sampling (CDS) technique [4]. However, the pixel cross-talk effect is still without an effective method to compensate the optical, electrical and architectural interferences simultaneously.

Contact IEEE to Subscribe

References

References is not available for this document.