Page 1 of 1
matching output from the identifty command.
Posted: 2015-11-23T07:40:57-07:00
by nitrolinux722
Hello,
I am using ImageMagick 6.9.2-6 Q16 x64 on WIndows 10. I recently use the "identify" command with the "verbose" switch to read some detailed information about an DDS image type. I'm trying to conform the information I received to be able to change another DDS image to the same attributes. The only difference I see between the two are the channel depth, RGB colors and Aplha channel. Below is the two differences:
unedited image:
Code: Select all
Channel depth:
red: 8-bit
green: 8-bit
blue: 8-bit
alpha: 8-bit
Channel statistics:
Pixels: 3130592
Red:
min: 0 (0)
max: 254 (0.996078)
mean: 96.6081 (0.378855)
standard deviation: 54.2584 (0.212778)
kurtosis: 0.35092
skewness: -0.383135
entropy: 0.70386
Green:
min: 0 (0)
max: 254 (0.996078)
mean: 81.2331 (0.318561)
standard deviation: 61.269 (0.240271)
kurtosis: -0.592613
skewness: 0.302419
entropy: 0.80537
Blue:
min: 0 (0)
max: 255 (1)
mean: 136.069 (0.533605)
standard deviation: 74.9801 (0.29404)
kurtosis: -0.0849212
skewness: -0.363942
entropy: 0.73598
Alpha:
min: 0 (0)
max: 255 (1)
mean: 181.715 (0.71261)
standard deviation: 115.291 (0.452123)
kurtosis: -1.11436
skewness: 0.940305
entropy: 0.114513
Image statistics:
Overall:
min: 0 (0)
max: 255 (1)
mean: 96.7988 (0.379603)
standard deviation: 80.0187 (0.313799)
kurtosis: -0.386124
skewness: 0.433003
entropy: 0.589931
Image I need to change to match unedited image:
Code: Select all
Channel depth:
red: 8-bit
green: 8-bit
blue: 8-bit
alpha: 1-bit
Channel statistics:
Pixels: 3130592
Red:
min: 0 (0)
max: 255 (1)
mean: 106.208 (0.416501)
standard deviation: 75.1977 (0.294893)
kurtosis: -0.985669
skewness: 0.152831
entropy: 0.898439
Green:
min: 0 (0)
max: 255 (1)
mean: 99.9096 (0.391803)
standard deviation: 74.6307 (0.29267)
kurtosis: -0.983288
skewness: 0.283349
entropy: 0.900504
Blue:
min: 0 (0)
max: 255 (1)
mean: 116.019 (0.454975)
standard deviation: 90.586 (0.355239)
kurtosis: -1.28064
skewness: 0.281044
entropy: 0.807191
Alpha:
min: 0 (0)
max: 255 (1)
mean: 211.305 (0.828648)
standard deviation: 96.0881 (0.376816)
kurtosis: 1.04274
skewness: 1.74434
entropy: 0.0825984
Image statistics:
Overall:
min: 0 (0)
max: 255 (1)
mean: 91.4577 (0.358658)
standard deviation: 84.651 (0.331965)
kurtosis: -0.606921
skewness: 0.627267
entropy: 0.672183
I've tried to make sense of the command-line tool "convert", but I am at a lost to how I would formulate a command to change the edited image to match the unedited image or even if it's possible. Any help would be appreciated.
-Jason
Re: matching output from the identifty command.
Posted: 2015-11-23T11:05:01-07:00
by fmw42
It is not likely you can match the two, since you are trying to change a 1-bit alpha channel to 8-bits. To do that you would need to blur or feather the alpha channel to anti-alias it to 8-bits.
If the two images are exactly the same data and size except for the alpha channel, then you can transfer the alpha channel from the one image to the other image.
Please post your images to some free hosting service such as dropbox.com and put the URL here so we can take look and see what might be done.
Re: matching output from the identifty command.
Posted: 2015-11-24T00:14:26-07:00
by nitrolinux722
Hello fmw42,
Thank you for the post. What I'm trying to do is modify a DDS file that is a part of a virtual screen background that we use to key out a green screen. I was able to get the image to finally to come up, but it had a purple tint. The image I'm using is one I created in Valve's Hammer. It's a way for me to create a 3D set without having to spend a thousand dollars on software that allows us to create these virtual sets. In my thinking, i could replace an image and try to match it close to the original attributes of the original image and maybe get some success. I'm kinda there and I wish i had an image of the success, but I don't at this moment. In this post I have included the files that are on my dropbox. Please take a look at them and tell me what you think.
https://www.dropbox.com/s/qksa069bvdod2 ... d.dds?dl=0 (unedited)
https://www.dropbox.com/s/sdvzuahqaqe9v ... d.dds?dl=0 (edited)
Thank you.
-Jason
Re: matching output from the identifty command.
Posted: 2015-11-24T10:07:51-07:00
by fmw42
Can you clarify what you are trying to do? Are you trying to take the bible sign from the one image and put it into the transparent area of the other image? Or are you trying to make the region of the bible sign transparent? Sorry I am confused about your goal.
Re: matching output from the identifty command.
Posted: 2015-11-24T18:25:27-07:00
by nitrolinux722
We have a product made by NewTek called tricaster 40. It uses what is called virtualsets. These images are like tv studios, but only virtual. The tricaster 40 uses these images, like the one DDS file that is unedited. Strangely enough, it does not appear that way once loaded into the tricaster. It's colors are correct. Not the purple tints like you see. Why I don't know. Maybe you can enlighten me on more about DDS file types. I know it belongs to Mircosoft. Direct Draw?? Now, the virtual set we are using now for our Bible Talk program is the studio107. We are trying to do something different for 2016. I was hoping to be able to conform my edited image like the original one, maybe giving it a chance to work. Then again, maybe it will be totally impossible to do this. On a side note, there are other files that come into play that makes this work. Below is an example of the directory structure:
/Studio107
| uv.dds
| ycbcra.dds
Studio107.LiveSet (an XML file)
Studio107.png
The Studio107.LiveSet file calls the DDS files to be loaded in the XML file. I'm sure I've confuse you greatly now, but if you have any ideal about this, please let me know.
Thank you,
-Jason
Re: matching output from the identifty command.
Posted: 2015-11-24T18:46:22-07:00
by fmw42
I still do not understand what you are trying to do with the two images. Can you provide a command line using the two files you provided? And explain the goal your are trying to achieve in terms of what you expect to see where in the edited image? The two images do not seem to conform to each other.
Does the unedited image display correctly in IM?
Re: matching output from the identifty command.
Posted: 2015-11-24T20:50:47-07:00
by nitrolinux722
I guess in simple terms I want to make the edited image like the unedited image when it comes to the red, blue, and green channels. I have not attempted to do any changes with the convert tool. I've pretty much resolved to the fact that number one, I really don't know how to ask what I'm trying to ask. And number two, I've wasted enough of your time. I'm just going to have to break down and purchase this . What is the best way to mark this thread? [GIVING UP]
LOL . Thank you for your help.
Re: matching output from the identifty command.
Posted: 2015-11-24T22:45:09-07:00
by fmw42
This seems to be about as good as I can do, though not very good.
I cropped your image and saved as png. Then did the following.
Code: Select all
convert ycbcra_unedited.png -set colorspace YCbCr -colorspace sRGB ycbcra_unedited_ycbcr2srgb.png
histmatch script or -recolor to the edited image just seems to make it worse.
Re: matching output from the identifty command.
Posted: 2015-11-25T00:08:14-07:00
by nitrolinux722
Thank you for your example. Is it possible to take the "color information", for the lack of the correct term, of one image and transfer that information to another image. Such as in these files. unedited image's RGB info to the edited image. To further explain, like in the very top post, I give the main differences of the two images. Like the Red channel of the unedited image is:
Red:
min: 0 (0)
max: 254 (0.996078)
mean: 96.6081 (0.378855)
standard deviation: 54.2584 (0.212778)
kurtosis: 0.35092
skewness: -0.383135
entropy: 0.70386
Make the edited image Red Channel match these same numbers. Is that possible?
-Jason
Re: matching output from the identifty command.
Posted: 2015-11-25T00:15:34-07:00
by fmw42
I tried that with my script histmatch, at the link below,
http://www.imagemagick.org/Usage/quantize/#remap (which matches histograms) and with the IM function -remap (see
http://www.imagemagick.org/Usage/quantize/#remap). Neither worked well. User snibgo also has a histogram matching technique at
http://im.snibgo.com/. Perhaps his approach will work better than mine.
But these work best when you have the same image content. So, if you can provide an example image before and after editing, rather than two different images, then perhaps we can validate some mechanism that reproduces that editing result.