exposure duration during video capture

roysalisbury

Using the ASIGetVideoData call for getting the next frame of video data, when is the "exposure duration" started? Is it from the last frame, or the start of the call to ASIGetVideoData.

For example... If I had

StartTimer()

ASIGetVideoData(...)

TotalExposureDuration = GetTimerElapsed()

- OR -

NextExposureDuration = ASIGetControlValue(ASI_EXPOSURE ...)

ASIGetVideoData(...)

Reason I ask is the following... Say I had set the exposure duration to 15 seconds. Then waited 5 seconds before calling ASIGetVideoData. Would the call take the full 15 seconds, or just the remaining 10 seconds.

And if the exposure time is set to 15 seconds, and I change the exposure duration to 1 second, does that stop the current exposure, rest the duration and then start a new capture?

The reason I am asking is that I am using "auto" for the duration and gain to its constantly updating and would like to know what the duration was for EACH frame.

If fact, it would be great if you could add a new API call (ASIGetVideoDataEx) that also returned the values of certain settings (exposure, gain, gamma) for each frame acquired. That way there is no mistake about what values were used.

Roy

chad

Hi Roy,

1, For long time exposure, I suggest you use snap mode.

2, For exposure duration, in you example, for 15s, delay 5s, then call getVideoData,

You's better wait 15s, the wait time should be longer than the exposure time to ensure the data can be received.

Usually, for 15s exposure, it needs 15.2 to 17s, exposure time need 15s, and transfer time need 0.2 to 2s.

3, GetVideoData's iWaitms.

It is a timeout time. We will wait iWaitms, during this time, if we receive the data, we will return, if not, we will return a timeour error.

4, ASIGetVideoDataEx.

It is a great idea, however, it is too hard to program. 1st, As we know, sometimes, the camera has a frame rate of hundreds of frames, it is hard to return every frame's parameters. 2nd, different camera have different control types, it is hard to return all of the types and the value in one funciton.

Thanks

Chad

roysalisbury

Its not the wait time I am looking for. Its the actual exposure time for the image returned (same goes for gain). I need to feed these images into a video encoder and need to know the duration of each image so I can get the frame rate correct.

I'm not using snap mode because I need the auto exposure/gain features of the continuous images. I just set a MAX exposure limit and let the camera do its work.

As for the suggested ASIGetVideoDataEx. But I still need to know the exposure duration and gain on each returned frame. If I don't know WHEN to get it (before the GetVideoData or after, then it should be returned. Just adding these two is all I am asking for .. most of the other values are not relevant (just the ones that can be auto). They are adjusted after each frame, so they are known somehow.

Roy

chad

Hi Roy,

Got your needs。

According to our design, it is difficult to get the current data's exposure time, if you use our auto-exp funciton.

Auto-Exp and get-Data are in 2 different thread. So they may be asynchronous.

Maybe you can program the auto-exp by yourself, then set it into our camera, and get the current frame data. It will be better.

I am sorry.

Thanks

Chad

roysalisbury

If you can give me the logic behind your auto exposure/gain settings I can give it a shot. However, it seems to me that would just add extra processing on the application side (going through the video data) which could result in more dropped frames.

But I'll give it a shot if you can send me the logic.

Roy

chad

Hi Roy,

There is a circle buffer to receive the data.

Capture thread only need to get the data from the sensor and put them into the buffer.

Auto-exp thread will analysis the newest data in the circle buffer and calcluate a new exposure time then set the new time into the camera.

So you see, it is asynchronous.

Thanks

Chad

roysalisbury

The background buffer would be possible, but what are the calculations on the images to set the next gain/ exposure values? I’m assuming your also using the background value in the calculation.

What calculations are used on the images themselves?

Roy

chad

Hi Roy,

Set a target brightness and then calculate the average brightness of the image. If the average brightness is greater than the target brightness, the exposure time will be reduced gradually. Otherwise the exposure time will be added gradually.

So it is not complicated. You also can program it in your side.

Thanks

Chad

roysalisbury

Perhaps I'll give it a try .. but offloading that calculation to the camera is better from a resource point of view.

Roy

chad

Hi Roy,

It is not what you think. This calculation is running on your PC, not the camera.

This calculation is in our SDK, so it is running in your PC. If it runs in the camera, the camera needs a ISP Chip. So far, we haven't found an ISP chip suitable for our astronomy application.

Beside, if you do it yourself, you can use a lot of functions from other libraries that are already very efficient, to avoid using our algorithm. Because our algorithm sacrifices some efficiency for stability.

Thanks

Chad

roysalisbury

OK.. Makes a bit of sense now why you recommended going to the single exposure vs the video.

What algorithm are you using for processing the image for brightness. A simple average of the pixels, an RMS (root mean square), or some other calculation. I'd like to start out by getting the results as close to the SDK version as possible and then optimize from there.

Other questions are what step increments do you use for increasing or decreasing the exposure and gain? Is it a simple scale based on how close to the brightness you are, or just something as simple as doubling or halving the last increment/decrement.

BTW, if I can get this working correctly, I will be more than happy to share the results with you for your C# examples. I've already optimized the bitmap creation for the 8 and 16 bit images. The loops you had were VERY slow.

Roy

chad

HI Roy，

there is no special about this.

We used mean, just counting some pixels, not all. But you can call other funciton in other librayies, like openCV. They will return you a mean value very fast.

For step, there are 2 ways, first, calculate the difference between the average value and the target value, and then multiply it by a coefficient. Determine the coefficient based on your debugging. This result can be used as a step. second, use other approximation algorithm, like successive approximation algorithm, to get the step.

Thanks

Chad

roysalisbury

I'll probably start on snapshot mode in the next few days once I wrap up a few more issues with the Video mode capture. I have a few algorithms that I think might work for the brightness calculations and will also check out the OpenCV libraries. I was just looking into that library to see how it deals with the image data (same pointer I get back from the SDK, or is it going to have to be converted to an actual bitmap first. Might be able to do some parallel processing of the data. The less I have to keep marshaling between managed and unmanaged code the better.

I also have to get the FFMPEG libraries plugged in to get an actual video stream and not just image frames. There is something you can add to the camera. An h264 video stream .. Then I would not have to write anything. :)

Roy

chad

Hi Roy，

Your idea is good.

But in fact, many of our customers believe that it is best to present the original data image. Because even a good compression will have an irreversible effect on image quality. They can do whatever it takes to get a higher image and video quality.

Thanks

Chad