Mixed resolution coding of stereoscopic video for Mobile devices

Abstract

This paper presents results for full and mixed resolution stereo video coding including subjective and objective evaluation for 3DTV on Mobile devices. For this, objective as well as subjective tests have been carried out with different sequences at different image sizes and evaluated on two different stereoscopic displays. The subjective tests showed, that for uncoded sequences full resolution was rated better, however for coded sequences at low bit rates mixed resolution was rated better. An unsharp masking algorithm for up-sampling the lower resolution view only partially improved the overall quality because coding artifacts were strengthened as well. Objective tests showed that the optimum bit rate distribution for mixed resolution stereo video coding is 30% to 35% of the total bit rate for the lower resolution view (half horizontal and half vertical). Finally, tests showed that mixed resolution stereo video decoding with up-sampling is less complex than decoding full resolution stereo video which is important for implementation on mobile devices

Subjective study on compressed asymmetric stereoscopic video

“…The impact of quantization was verified in [4], which concluded that the perceived quality of coded equal-resolution stereo image pairs was approximately the average of the perceived qualities of the high-quality image and the low-quality image of the stereo pairs. A subjective evaluation for fulland mixed-resolution stereo video coding on a 32-inch polarization stereo display and on a 3.5-inch mobile display was presented in [5]. The results revealed that fullresolution (FR) was rated better for uncoded sequences in 79% of the test cases, while MR where one of the views was downsampled to a quarter resolution was rated better in 57% of the test cases for H.264/AVC-coded sequences.…”

Impact of downsampling ratio in mixed-resolution stereoscopic video

“…Dancer 960x540 480x270 360x202 240x135 Dog 800x600 400x300 300x225 200x150 Pantomime 800x600 400x300 300x225 200x150 Champagne 800x600 400x300 300x225 200x150 Newspaper 768x576 384x288 288x216 192x144 experiment was generally higher than in [3]. The exact values for pixels per viewing angle could not be concluded from the information given in [4], but it discovered equivalently the subjective difference between FR and MR was a descending function of the viewing distance or the resolution in terms of number of pixels. We also analyzed whether the subjective image quality ratings had a correlation to the average luma peak signal-to-noise ratio (PSNR) of the lower resolution view.…”

“…The study presented in [4] included a subjective evaluation for full-and mixed-resolution stereo video on a 32-inch polarization stereo display and on a 3.5-inch mobile display. One of the views in the MR sequences was downsampled to a half resolution both horizontally and vertically.…”

Development of a new MPEG standard for advanced 3D video applications

“…Taking the bitrate for the left view as given for 2D video, the 3D video functionality could be added by an overhead of 25-30% for coding the right view at quarter resolution. Such and similar approaches are currently under investigation [5]. A general drawback of CSV is that the 3D impression cannot be modified.…”