Resolution and Bit Depth Are Different
Our inspection camera captured the metal part, but the scratch is faint and partly hidden in shadow.
Resolution and Bit Depth Are Different explained visually
Our inspection camera captured the metal part, but the scratch is faint and partly hidden in shadow. English captions are optional and start off. The complete transcript follows the written lesson.
What you will learn
Our inspection camera captured the metal part, but the scratch is faint and partly hidden in shadow.
The faint scratch
The same metal part is back under the camera, but its scratch is hard to see in the shadow. Would more pixels help, or would more intensity levels help? Those sound like the same upgrade, but they aren't. Let's separate the two before changing our capture settings.
Spatial resolution
Resolution tells us how many locations the camera sampled. Start with an image that is six hundred and forty by four hundred and eighty pixels. Shrink it to three hundred and twenty by two hundred and forty, and each side now has half as many locations. The total falls from three hundred and seven thousand, two hundred to seventy-six thousand, eight hundred. A very narrow scratch might disappear. Repeating fine lines can also turn into false wider patterns; that's aliasing. Filtering before the resize can reduce it, but enlarging the smaller image later won't bring back measurements that were never kept.
Intensity levels
Bit depth is about the possible values at each location. An eight-bit channel has two hundred and fifty-six integer levels; a sixteen-bit channel has sixty-five thousand, five hundred and thirty-six. More levels can help preserve subtle differences in a shadow if the sensor actually captured them. They cannot reveal a scratch that is hidden or out of focus.
Choosing a capture
There is a storage trade-off as well. A hundred-by-hundred grayscale array stores ten thousand pixel values: ten thousand bytes at eight bits, or twenty thousand at sixteen bits, before headers and compression. So ask what information the inspection needs, and preserve the original. Next, we'll put color channels into this picture.
Worked example
An uncompressed 100-by-100 grayscale array contains 10,000 samples. At eight bits per sample, its pixel data needs 10,000 bytes. At sixteen bits, it needs 20,000 bytes. File headers and compression can change the actual file size.
Quick check: test your understanding
Try it yourself: If both dimensions double while bit depth stays fixed, how does uncompressed pixel storage change?
Check the reasoning
It becomes four times larger: twice as many rows multiplied by twice as many columns.
Does converting an eight-bit image to sixteen bits recover missing shades?
No. The larger container holds the existing values more precisely but does not add lost sensor information.
How many locations after halving both dimensions of a 640 × 480 image?
76,800 locations, one quarter of the original.
Does upscaling recover discarded measurements?
No. It adds pixels but not lost samples.
Further reading
Digital Image Processing, 4th edition — Rafael C. Gonzalez and Richard E. Woods.
Original ToolsInVision teaching examples. Editorial approach · Suggest a correction
Accessible referenceLesson 2 transcript257 words · Video transcript · Matches the published narration and captionsOpen transcript
The same metal part is back under the camera, but its scratch is hard to see in the shadow. Would more pixels help, or would more intensity levels help? Those sound like the same upgrade, but they aren't. Let's separate the two before changing our capture settings.
Resolution tells us how many locations the camera sampled. Start with an image that is six hundred and forty by four hundred and eighty pixels. Shrink it to three hundred and twenty by two hundred and forty, and each side now has half as many locations. The total falls from three hundred and seven thousand, two hundred to seventy-six thousand, eight hundred. A very narrow scratch might disappear. Repeating fine lines can also turn into false wider patterns; that's aliasing. Filtering before the resize can reduce it, but enlarging the smaller image later won't bring back measurements that were never kept.
Bit depth is about the possible values at each location. An eight-bit channel has two hundred and fifty-six integer levels; a sixteen-bit channel has sixty-five thousand, five hundred and thirty-six. More levels can help preserve subtle differences in a shadow if the sensor actually captured them. They cannot reveal a scratch that is hidden or out of focus.
There is a storage trade-off as well. A hundred-by-hundred grayscale array stores ten thousand pixel values: ten thousand bytes at eight bits, or twenty thousand at sixteen bits, before headers and compression. So ask what information the inspection needs, and preserve the original. Next, we'll put color channels into this picture.