The problematic is which resolution to choose when ripping a DVD for a given bitrate. The smaller the resolution, the better the encoded video will be. But when played in fullscreen mode, the differences with the reference video will be more magnified. On the contrary, the larger the resolution is, the worst its quality will be. But, it will less suffer from being played in fullscreen mode.
As a reference video, I used the Special Edition DVD of Starship Troopers. I encoded two videos with MeGui using the x264 codec. Their resolutions are 720x392 and 1072x576. The bitrate is 1000 kilobits per second. The Avisynth scripts used to encode these videos are :
DGDecode_mpeg2source("D:\STARSHIP_TROOPERS\VTS_01_1.d2v",info=3)
ColorMatrix(hints=true)
crop( 8, 8, -6, -14)
BilinearResize(720,392)
and
DGDecode_mpeg2source("D:\STARSHIP_TROOPERS\VTS_01_1.d2v",info=3)
ColorMatrix(hints=true)
crop( 8, 8, -6, -14)
BicubicResize(1072,576,0,0.5) # Bicubic (Neutral)
I use Avisynth to compare the encoded videos with the DVD source.
LoadPlugin("Compare2.dll")
# DVD source resized
DGDecode_mpeg2source("D:\STARSHIP_TROOPERS\VTS_01_1.d2v",info=3)
ColorMatrix(hints=true)
crop( 8, 8, -6, -14)
Clip1=BicubicResize(1600,864,0,0.5)
Clip2=AVCSource("..\Starship_Troopers.dga").BicubicResize(1600,864,0,0.5)
#Clip2=AVCSource("..\Starship_Troopers 2.dga").BicubicResize(1600,864,0,0.5)
Compare2(Clip1,Clip2,"YUV","720-392.log",false)
#Compare2(Clip1,Clip2,"YUV","1072-576.log",false)
As you can see, I assume the videos would be played in fullscreen mode on a 1600-pixel width screen. I use my own modified version of Avisynth plug-in Compare to measure two metrics : PSNR and SSIM.
Total frames processed : 186459
| PSNR(dB) | ||||
| Channel | Minimum | Maximum | Average | Overall |
| Y | 24.473 | 100.000 | 40.711 | 39.268 |
| U | 34.993 | 100.000 | 48.211 | 46.371 |
| V | 36.308 | 100.000 | 48.925 | 46.966 |
| YUV | 26.184 | 100.000 | 41.919 | 40.650 |
| Scaled overall YUV SSIM | |
| SSIM | 82.280 |
| SSIM luma1 | 79.771 |
| SSIM luma2 | 79.754 |
Total frames processed : 186459
| PSNR(dB) | ||||
| Channel | Minimum | Maximum | Average | Overall |
| Y | 28.084 | 100.000 | 41.771 | 40.890 |
| U | 36.966 | 100.000 | 47.874 | 46.141 |
| V | 37.582 | 100.000 | 48.556 | 46.658 |
| YUV | 29.752 | 100.000 | 42.869 | 42.079 |
| Scaled overall YUV SSIM | |
| SSIM | 82.573 |
| SSIM luma1 | 80.194 |
| SSIM luma2 | 80.171 |
First, remember that the two video files have the same length since they have the same bitrate.
As you can see, there is little difference between the two encoded videos when they are compared with reference video in "fullscreen mode". Results are slightly better with the larger resolution except with the chroma channels (U and V) according to the PSNR metric.
So, there is little difference in the end, but it takes to encode at 1072x576 approximately twice the time to encode at 720x392.
Eventually, if you really want the best quality, you should go with the higher resolution. Yet, considering the few differences and the gain in time, you can pick the lower resolution.