Explanation of arguments to extract_centroids()

[windell747]

Thank you so much for your help!

1. Here is a copy and paste from cedar_detect.proto. My questions are located next to each variable. Im wondering if you could comment? Im havign trouble understanding some of the comments in the .proto file and im wondering if you might be able to explain them better here.

double sigma = 2; //I see the commend in the .proto. So I think this is the SNR of the pixel threshold?

int32 max_size = 3 [deprecated=true]; //This is the maximum amount of pixels containing in a circular area of a valid star? So if the diameter in an image is approximately 8 pixels, then the max_size could be approximately 50 pixels assuming circular sources?

optional int32 binning = 8; //so this should be either 2 or 4 meaning 2x2 or 4x4 binning on the original image? How does plate scale get affected? What do I do if i dont want binning?

bool return_binned = 4; //Im not sure what this argument means.

bool use_binned_for_star_candidates = 5; //I see the comment in the code, but im still not sure waht this means

// If true, hot pixels are detected (prior to binning, if used) and do not
// contribute to detected stars. If false, hot pixels could be falsely
// detected as stars.
bool detect_hot_pixels = 6; //im guessing this means that hot/saturated pixels will be used in the centroiding if this is false?

// If provided, specifies a portion of input_image over which the background
// level (excluding stars) is estimated.
optional Rectangle estimate_background_region = 7;

2. Some of these variables are bool, but they aren't 0 or 1. Is there a reason for this? Im assuming anything >=1 is considered true?
3. Is there a way to see what the grid size that is used to determine the SNR for a particular centroid?

[smroid]

Thank you so much for your help!

1. Here is a copy and paste from cedar_detect.proto. My questions are located next to each variable. Im wondering if you could comment? Im havign trouble understanding some of the comments in the .proto file and im wondering if you might be able to explain them better here.

double sigma = 2; //I see the commend in the .proto. So I think this is the SNR of the pixel threshold?

Yes. Set this to 8 as a starting point; I've never deviated far from that.

int32 max_size = 3 [deprecated=true]; //This is the maximum amount of pixels containing in a circular area of a valid star? So if the diameter in an image is approximately 8 pixels, then the max_size could be approximately 50 pixels assuming circular sources?

max_size field is ignored now.

optional int32 binning = 8; //so this should be either 2 or 4 meaning 2x2 or 4x4 binning on the original image? How does plate scale get affected? What do I do if i dont want binning?

Plate scale is unaffected. The StarCentroid.centroid_position is reported in full resolution coordinates regardless of the binning value.

If you don't want binning, set false for use_binned_for_star_candidates (default).

bool return_binned = 4; //Im not sure what this argument means.

CentroidsResult will contain 'binned_image' if return_binned is true.

Rationale: use_binned_for_star_candidates is typically true; separately apps might want the binned image. This way we compute the binned image once-- for star detection, and then to be returned to the app.

bool use_binned_for_star_candidates = 5; //I see the comment in the code, but im still not sure waht this means

This controls whether star detection happens in the full-res image or on a 2x or 4x binned image.

// If true, hot pixels are detected (prior to binning, if used) and do not // contribute to detected stars. If false, hot pixels could be falsely // detected as stars. bool detect_hot_pixels = 6; //im guessing this means that hot/saturated pixels will be used in the centroiding if this is false?

Correct, if false a hot pixel will be reported as a star. If true hot pixels are detected and skipped.

// If provided, specifies a portion of input_image over which the background // level (excluding stars) is estimated. optional Rectangle estimate_background_region = 7;

2. Some of these variables are bool, but they aren't 0 or 1. Is there a reason for this? Im assuming anything >=1 is considered true?

In protocol buffer, a bool field is either true or false, defaulting to false.

3. Is there a way to see what the grid size that is used to determine the SNR for a particular centroid?

There is not currently a way to find out what portion of the input image is used for noise estimation.

[windell747]

Thank you very much! Is there any way to enable use of max_size? Or is it not implemented in the code? What happens if the center of the centroid (star candidate) is saturated and the detect_hot_pixels is enabled. Is just that pixel not included in the calculation or is the whole centroid rejected? What does the estimate_background_region do? Sorry, I dont think i see the explanation in your previous response. Again thank you very much!

…

On Thu, Mar 6, 2025 at 6:06 PM Steven Rosenthal ***@***.***> wrote: Thank you so much for your help! 1. Here is a copy and paste from cedar_detect.proto. My questions are located next to each variable. Im wondering if you could comment? Im havign trouble understanding some of the comments in the .proto file and im wondering if you might be able to explain them better here. double sigma = 2; //I see the commend in the .proto. So I think this is the SNR of the pixel threshold? Yes. Set this to 8 as a starting point; I've never deviated far from that. int32 max_size = 3 [deprecated=true]; //This is the maximum amount of pixels containing in a circular area of a valid star? So if the diameter in an image is approximately 8 pixels, then the max_size could be approximately 50 pixels assuming circular sources? max_size field is ignored now. optional int32 binning = 8; //so this should be either 2 or 4 meaning 2x2 or 4x4 binning on the original image? How does plate scale get affected? What do I do if i dont want binning? Plate scale is unaffected. The StarCentroid.centroid_position is reported in full resolution coordinates regardless of the binning value. If you don't want binning, set false for use_binned_for_star_candidates (default). bool return_binned = 4; //Im not sure what this argument means. CentroidsResult will contain 'binned_image' if return_binned is true. Rationale: use_binned_for_star_candidates is typically true; separately apps might want the binned image. This way we compute the binned image once-- for star detection, and then to be returned to the app. bool use_binned_for_star_candidates = 5; //I see the comment in the code, but im still not sure waht this means This controls whether star detection happens in the full-res image or on a 2x or 4x binned image. // If true, hot pixels are detected (prior to binning, if used) and do not // contribute to detected stars. If false, hot pixels could be falsely // detected as stars. bool detect_hot_pixels = 6; //im guessing this means that hot/saturated pixels will be used in the centroiding if this is false? Correct, if false a hot pixel will be reported as a star. If true hot pixels are detected and skipped. // If provided, specifies a portion of input_image over which the background // level (excluding stars) is estimated. optional Rectangle estimate_background_region = 7; 2. Some of these variables are bool, but they aren't 0 or 1. Is there a reason for this? Im assuming anything >=1 is considered true? In protocol buffer, a bool field is either true or false, defaulting to false. 3. Is there a way to see what the grid size that is used to determine the SNR for a particular centroid? There is not currently a way to find out what portion of the input image is used for noise estimation. — Reply to this email directly, view it on GitHub <#7 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHZ3GGRW2FTWFYPBMDLELTT2TELKTAVCNFSM6AAAAABYQHAU2KVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTENBSGE2TMMY> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[smroid]

Thank you very much! Is there any way to enable use of max_size? Or is it not implemented in the code?

The max size logic is now hardwired, no longer can set the parameter.

What happens if the center of the centroid (star candidate) is saturated and the detect_hot_pixels is enabled. Is just that pixel not included in the calculation or is the whole centroid rejected?

detect_hot_pixels detects single isolated bright pixels and replaces them with the mean of the two neighbors (left and right). So if the center of a centroid is saturated, but surrounding pixels are also bright, the hot pixel detection will not trigger.

What does the estimate_background_region do? Sorry, I dont think i see the explanation in your previous response.

estimate_background_region allows you to specify a portion of the input image over which an estimate of the background (with stars removed) is made. This is intended to support things like SQM computation.