Frequency: Another Way to Describe an Image
An image of a smooth sky changes slowly across space.
Frequency: Another Way to Describe an Image explained visually
An image of a smooth sky changes slowly across space. English captions are optional and start off. The complete transcript follows the written lesson.
What you will learn
An image of a smooth sky changes slowly across space.
Separate slow and rapid changes
A fine repeating stripe appears on our inspection image. Rather than following each stripe, ask how often the pattern repeats across the frame. Broad bands change slowly with position; a tight checkerboard changes rapidly. Spatial frequency describes those rates and their directions.
Filter selected patterns
A Fourier transform represents the image with components at different frequencies. Its magnitude tells us how strongly each component contributes, while phase helps determine where structures line up. For viewing, software often moves the slowest component to the center of a plot. That's a display choice, not filtering.
Remember the tradeoff
Compare the stripe image with its transform. A low-pass filter weakens rapid change and may smooth both stripe noise and real scratches. A high-pass filter emphasizes edges and grain. If a periodic artifact produces narrow peaks, a selective filter may help, but genuine repeated texture can live nearby. Abrupt frequency cuts can also make ringing near edges.
A worked example
Keep the untouched part beside every filtered result. We have another view of the same evidence, not new evidence. Next, we'll ask whether a blurred or damaged image can be restored faithfully at all.
Worked example
Suppose a photograph has evenly spaced bright stripes caused by interference. Their regular spacing creates a concentrated periodic pattern. A selective frequency filter may reduce the stripes, but removing nearby useful frequencies can also damage real textures with similar spacing.
Quick check: test your understanding
Try it yourself: Name one useful image feature and one unwanted feature that a high-pass filter might both emphasize.
Check the reasoning
It might emphasize a real object edge and also fine sensor noise. Frequency alone does not distinguish useful detail from unwanted variation.
Can phase be ignored when reconstructing an image accurately?
No. Phase contributes essential spatial alignment information.
How would you solve the practice case?
It might emphasize a real object edge and also fine sensor noise. Frequency alone does not distinguish useful detail from unwanted variation.
What does the worked example show?
Suppose a photograph has evenly spaced bright stripes caused by interference. Their regular spacing creates a concentrated periodic pattern. A selective frequency filter may reduce the stripes, but removing nearby useful frequencies can also damage real textures with similar spacing.
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 13 transcript181 words · Video transcript · Matches the published narration and captionsOpen transcript
A fine repeating stripe appears on our inspection image. Rather than following each stripe, ask how often the pattern repeats across the frame. Broad bands change slowly with position; a tight checkerboard changes rapidly. Spatial frequency describes those rates and their directions.
A Fourier transform represents the image with components at different frequencies. Its magnitude tells us how strongly each component contributes, while phase helps determine where structures line up. For viewing, software often moves the slowest component to the center of a plot. That's a display choice, not filtering.
Compare the stripe image with its transform. A low-pass filter weakens rapid change and may smooth both stripe noise and real scratches. A high-pass filter emphasizes edges and grain. If a periodic artifact produces narrow peaks, a selective filter may help, but genuine repeated texture can live nearby. Abrupt frequency cuts can also make ringing near edges.
Keep the untouched part beside every filtered result. We have another view of the same evidence, not new evidence. Next, we'll ask whether a blurred or damaged image can be restored faithfully at all.