Processing Decoded Video for Backlight Dimming: Video Quality Enhancement on LCD with Dynamic Local Backlight

Quality of digital image and video signals on TV screens is aected by many factors, including the display technology and compression standards. An accurate knowledge of the characteristics of the display andof the video signals can be used to develop advanced algorithms that improve the visual rendition of the signals, particularly in the case of LCDs with dynamic local backlight. This thesis shows that it is possible to model LCDs with dynamic backlight to design algorithms that improve the visual quality of 2D and 3D content, and that digital video coding artifacts like blocking or ringing can be reduced with post-processing. LCD screens with dynamic local backlight are modeled in their main aspects, like pixel luminance, light diusion and light perception. Following the model, novel algorithms based on optimization are presented and extended, then reduced in complexity, to produce backlights that find optimal balance between image quality and power consumption. The algorithms are tested in several experiments in comparison with other State of the Art approaches.
Based on the model of backlight dimming, another model of backlight scanning for crosstalk reduction in time-sequential stereoscopic visualization on LCD is introduced. Crosstalk at a given luminance level is minimized by properly adjusting the signals controlling the backlight.