Of course, for better accuracy you should locate your point at higher resolution. However, I do not know what are your images, how large and how smeared are your points and what are your reasons to do it this way. At downsampling, the software (e.g. Photoshop) will try to keep the point position at the same relative coordinates by appropriate aliasing. You can estimate the point position by your visual perception, or to use some appropriate software tool to locate the peak position.

If you do geometric spatial transformation using MATLAB "for example"; the ratios of distances between pixels will be preserved and "by default" Bilinear interpolation will be achieved for those new "empty" pixels.

But if you do a direct interpolation using the instruction "interp2(Your_Image)" the dimensions of the image will be doubled in this case "2*original_size - 1". That means you have the original pixels and nearly the same number of interpolated pixels.

Now, any down-sampling step (Manually or using Wavelets) will not affect the relative coordinates of pixels. But you have to note, that your samples will be taken from the upper-mentioned doubled size image with all allowed pixels (original and interpolated).

hi wadha,

sure the location will be changed and in a simple calculation you can multiply the coordinated by the sizing ratio, in your example the sizing ratio is 200/1000 = 0.2, then you need to multiply the coordinates by 0-.2

best,

Mohammad

I can't answer, you don't tell the initial image resolution ! And you don't tell how you measure the point !

Very interesting question. With not an easy answer. You should test by yourself how the scaled image affect your point recognition process.

Hi,

The image resizing impact depends on several factors such as:

- Your original image (size, resolution, contrast etc)

- Your localization algorithm

- The interpolation algorithm

- The image format or pattern:

- Do not use interpolated JPG, JPEG. It may create some undesired artifacts in the image.

Generally, if you interpolate the image to be 1000 by 1000 pixels, then downsample it to 200 *200 pixels, you will lose information about the original point position. You can test the tradeoff between the algorithm speed and the algorithm accuracy changing your parameters (new image size, downsample rate etc).

Instead of increasing the size of the image using an interpolation algorithm, you can try to acquire an image with a better resolution (if it is possible). Did you test your algorithm using:

- Your original image?

- Your original image with interpolation and without downsampling?

- Your original image without interpolation and downsampled?

Best Regards,

Teruo

Yes, of course , the location will be changed. as mentioned by Mohammad Al-Azawi you can get the relative position by doing that simple math. And you may not see the actual intensity value in that location.