The results for MPEG-1 encoding

Below we show a comparative study on one video clip, looking at all combinations of the model parameters described above.

• Video compression was performed using the mpeg_encode program.
• The original clip encoded as MPEG-1 has a file size of 10.8 MB (11317048 bytes).
• The original clip was captured using an Hi8 analog video camera, then played on the camera and digitized from the interlaced composite output of the camera, using a consumer-electronics frame grabber (Hauppage WinTV Go), at 640x480 resolution, YUV9 color and 30 frames/s. The clip has 481 frames.
• The MPEG-1 encoder settings were identical for all clips shown on this page.
• In the table below, MPEG-1 encoded size ratios are shown for each variation of the saliency-based multi-foveation algorithm, that is, multi-foveated then MPEG-1 encoded size divided by original MPEG-1 encoded size times 100.
• Click on the size links (e.g., 27.2%) to see the resulting clip, and on the corresponding 'demo' link to see the associated combination demo clip. The bargraphs are only a visual rendition of the percentage numbers (full green is 0%, i.e., maximal gain from our multi-foveation technique, and full red is 100%, i.e., no gain from our multi-foveation technique)

Windows users beware: Somehow, MPEG-1 clips encoded with the reference program mpeg_encode do not play smoothly on Windows 2000 using the Windows Media Player anymore (they used to, but new versions of the Windows Media Player have broken that). They play well on Windows 98 and Windows NT (at least the versions we have here). If the movies appear choppy (playing too fast for a second, then pausing, then continuing too fast, etc), we recommend that you try playing them with the Quicktime Player rather than the Windows Media Player. All movies play smoothly at 30 frames/s under Linux using mpeg_play or mplayer.

Saliency Competition	Object Segmentation	Pyramid Depth 2				Pyramid Depth 4				Pyramid Depth 6
		Continuous	1 Fovea	3 Foveas	5 Foveas	Continuous	1 Fovea	3 Foveas	5 Foveas	Continuous	1 Fovea	3 Foveas	5 Foveas
Maxnorm	None	64.2% demo	53.2% demo	60.3% demo	63.9% demo	37.5% demo	31.5% demo	38.7% demo	42.7% demo	30.2% demo	25.6% demo	32.5% demo	36.4% demo
	FeatureMap		71.6% demo	87.4% demo	93.9% demo		51% demo	74.6% demo	86.5% demo		44.4% demo	69.5% demo	83.1% demo
FancyOne	None	42.2% demo	53.2% demo	61.6% demo	65.8% demo	27.2% demo	31.3% demo	39.5% demo	44% demo	23.6% demo	25.4% demo	33.3% demo	37.6% demo
	FeatureMap		60% demo	73.3% demo	78.7% demo		38.6% demo	52.9% demo	60.4% demo		32.4% demo	46.3% demo	53.9% demo
Fancy	None	35.9% demo	53.7% demo	62.9% demo	65.6% demo	20.6% demo	31% demo	40.3% demo	43.7% demo	17% demo	25.3% demo	33.7% demo	37.1% demo
	FeatureMap		57.1% demo	68.7% demo	71% demo		34.6% demo	46.9% demo	49.7% demo		28.6% demo	40.1% demo	42.9% demo

Results with I-Frame optimization

We here explore one additional variation, in which the blur mask is only allowed to move on frames whose number is a multiple of 3. Given our encoder settings, this means that the blur mask will move only on I-Frames or P-Frames of the MPEG-1 sequence. We measured virtually no improvement in file size using this technique, but great loss in visual quality of the results. We hence recommend not to use this additional optimization, at least for clips like this one which have significant full-field motion.

Saliency Competition	Object Segmentation	Pyramid Depth 2				Pyramid Depth 4				Pyramid Depth 6
		Continuous	1 Fovea	3 Foveas	5 Foveas	Continuous	1 Fovea	3 Foveas	5 Foveas	Continuous	1 Fovea	3 Foveas	5 Foveas
Maxnorm	None	64.1% demo	53.2% demo	60.3% demo	63.9% demo	37.4% demo	31.5% demo	38.7% demo	42.7% demo	30.2% demo	25.5% demo	32.3% demo	36.3% demo
	FeatureMap		71.6% demo	87.4% demo	93.9% demo		50.9% demo	74.6% demo	86.5% demo		44.1% demo	69.3% demo	83.1% demo
FancyOne	None	42.1% demo	53.2% demo	61.6% demo	65.8% demo	27.2% demo	31.3% demo	39.6% demo	44% demo	23.6% demo	25.3% demo	33.2% demo	37.5% demo
	FeatureMap		60% demo	73.3% demo	78.7% demo		38.6% demo	53% demo	60.5% demo		32.3% demo	46.3% demo	53.9% demo
Fancy	None	35.8% demo	53.7% demo	62.9% demo	65.6% demo	20.6% demo	31.1% demo	40.3% demo	43.7% demo	17% demo	25.1% demo	33.6% demo	37.1% demo
	FeatureMap		57.1% demo	68.6% demo	71% demo		34.6% demo	46.9% demo	49.7% demo		28.5% demo	40% demo	42.7% demo