I tried the SizingImg program. It works but I didn't get the quality I was looking for in the output image. It does shrink the image but the image was a little too blurry for my taste.
Perhaps, I just don't know how to use the right combo of params. I'll keep playing with it.
Not sure if there is a trial version. The site says it is shareware in that if you like it enough, you can send $ to the author.
I see no limitations on the application. All it is is a single exe file (about 80kb I think). :)
I was resizing an image about 850px in height to half with quality set to default at 75%. When I forced quality to 100%, the image quality is retained. The file size however didn't change that much. I lowered the quality to 90% and it seems to reduce the file size a bit better with an almost indiscernible dropoff in quality.
I'll play with it a bit more.
Quote: Originally posted by zxcvbnm on February-02-2006
Quote: . Check out www.jpeg.com they do a program that is superior for making the smallest file possible.
Hmm, I glanced at www.jpeg.com and got a porn advert.
Hmm, that must have been an added bonus! I have amended the link. Try www.jpg.com
Subject: Sizing down, reply to Wide-Eyes
As you know I have reputation of writing long posts. So be it.
I try to explain the basis of things as well as "how to do it" in a very simple way so that it can be understood by everyone.
The number of steps depends on the amount of sizing down needed.
Consider the following:
Scenario 1 = 4000 x 2000 sized down to 500 x 250
Scenario 2 = 4000 x 2000 sized down to 3000 x 1500
Scenario 1 requires a reduction of 8 times the original
Scenario 2 requires a reduction of 1.3 times the original
When you reduce the dimensional size of an image you are removing pixels. This results in "interpolation". Photoshop attempts to re blend the image by reducing pixels in an orderly way. A good example of this is when you make a GIF file. You can choose the # of colors to control the image size. Look at the difference between a GIF at 64 colors vs one at 256 colors. IF you convert a nice JPEG image to a gif and preview the quality in the "save for web" command you can immediately see the difference this makes in the image. One with 64 colors will look grainy and one with 256 will look quite good. This is because when reducing colors you are reducing the "tonal range" available to the image. If there is an object with a solid color this color may actually contain many variations in tone of that color. So reducing the # of colors by half will take out half of the tonal range in the object. The pixels that once blended nicely into the next pixel now have a greater difference in color. If you look at the color pallet in photoshop and choose "web colors" you see how this works. Change it back to all colors and you see a very smooth graduation of color. In web colors you see big blocks of color.
This is the same concept that happens in sizing down except that you are not removing the number of colorsyou are removing "pixels" of "every" color. The effect of this "interpolation" is exactly like "anti aliasing". Photoshop attempts the re alias the edges of sharp difference in color that results from removing pixels. It effectively tries to "re blend" or "re make" the tonal range. This results in the image becoming softer. In fact this effects different parts of the image more than others depending on the tonal range of color in that area. For example if the area in question is completely white you won't see any difference in that area when sizing down because all the pixels are white and therefore removing white from white will result in a smaller area of white. No difference.
It is areas of the image that have many colors that suffer the most. A typical "room" will have all kinds of things that have patterns of many colors like a bedspread or chair. These areas of the image will suffer more than areas of solid color. Especially in the areas that meet each other with totally different colors. Like the edge of a bedspread meeting the window or a chair with a weaved/ textured fabric. These are the areas where the anti aliasing takes place. This is also what makes the image look "softer"
When you "sharpen" an image photoshop scans the image and looks for areas of sharp definition (wide range of colors meeting a solid color). Say the arm of a wood chair that has the carpet in the background or a shadow of a texture like the ceiling. Photoshop will find these areas and replace certain pixels with black. This makes the definition of color appear sharper because the blending between these areas is less or "sharper" (for example - the difference between anti aliased text and non aliased text)
By sizing down in steps you are removing a smaller number of pixels each time and therefore reducing the amount of interpolation. To correct this you then sharpen the image in a small amount as well. So we end up with an image that looks as sharp as the original but a little smaller. We then size down again and sharpen. This allows the image to be sized down in a more orderly fashion.
Taking the image down in one big whack will introduce a huge amount of interpolation. When you attempt to correct this with one big sharpen you end up with an image that has very big tonal difference at the edges of different color because you are introducing a lot of black pixels all at once instead of having these progress down with the image size. Throw in some jpeg compression and you have a mess. The "mess" gets worse the more you have to size down. If you are sizing down a small amount this effect is negligible.
Now to the amount steps required.
Scenario 1 will require more steps than scenario 2. I try to size down in about 20% to 30% steps if the image will be sized down to more than 50% of its original size. This results in 3 to 5 steps.
For scenario 2, I would use two steps. First I would apply a sharpen of 25% to 50% to the original image (try 25% first I think that is best) Step one would be to 3500 x 1750 and sharpen 15%. Step two would be to 3000 x 1500 and another sharpen of 15%.
By adding a good amount of sharpen to the orginal image first you are adding these black pixels in preparation of the image being sized down. When the sizing down takes place these black pixels are included in the interpolation and "some of the sharpness" is retained at the smaller size. This continues to the final size. At the end you look at the image and decide if you need a final "touch up" sharpen of a very small amount - about 10%.
This business of sizing down is all about colors and not DPI or resolution. It makes no difference if the image has 300 dpi or 72 dpi. Actually its Pixels per inch or PPI to be correct. DPI really refers to a display device such as a monitor or printer.
PPI effects the "dimensional size" of the image on a monitor and not the sharpness. The sharpness is all about the difference in color definition between colors.
I simply recorded an action in photo shop to do the step sizing. This is much faster. I have made several with different steps and amount of sharpening. So I can run the action and if the image is not like I want I undo and run a different action.
Finally you have to consider your display size. You need to take this into account when deciding on the # of steps for sizing down. This gets into the subject of "angular resolution" and this is related to the DPI of the monitor. Not actually related to this post so I will end here.
Begin Wide-Eyes quote:
1) is it so that the more steps you take when rezising the better?? would 5 steps from 4000x2000 to 2000x1000 be better then 3 steps??
2) @ gen lee. Why do you start at 50% unsharp?? and not e.g. 75%?? And why do you go down to 15% unsharp the next couple of steps?? I assume it is to keep the quality as good a possible, but honestly I can't see the difference in my images after just 1 step, that is why I ask. I just want to be sure that I am doing it the best way :o)
First, the short, "how to do it answer"
Use a larger source image of 4000 x2000 or larger. Start testing by using 10% compression and continue in 10% increments up to 40%. If this does not get you to your target file size resize the source image a little smaller to about say 3500 x 1750 and start again. The concept is to use the largest size source image possible with more compression.
Using a large source file and more compression will result in better quality at your target file size than starting with a smaller source image and using less compression.
This is basically what Smooth said in his earlier post.
If you care to learn a little more here is the long answer with details:
I recommend you start with at least a 4000 x 2000 image and for even better quality start with 5000x2500.
Panoweaver is a great stitching program but it does not handle sizing and compression as good and you cannot control cube face size.
So we must use a different method for this.
I use "Pano2QTVR" to make my full screen QTVR's. It is free to use. You get added features by paying for it. Like adding sprites to your QTVR for controlling it with auto pan, adding sound tracks and so forth. If you don't need these features you can use it to simply make the QTVR mov file and display it however you want.
A QTVR is constructed of 6 "cube faces" that are created from an equirectangular image (your stitched pano). This image is converted to 6 corrected flat images. They are assembled into a box (cube) and displayed by Quicktime as a movie.
We start with the original stitched image at 4000 x 2000
We then select a cube face size. It is the cube face size that effects the final file size in bytes.
After the image is created we will then select a compression amount.
The result will be your QTVR or your ".mov" file.
There are two ways to arrive at a target file size in bytes. Lets say the target size is 1mb for a full screen image.
#1 - Start with a smaller equirectangular image.say 2000 x 1000 and use a small amount of compression.say 10% or none at all
#2 - Start with a larger equirectangular image.say 4000 x 2000 and use a much larger amount of compression..say 40%
Both of these methods will result in getting close to your file size but in totally different ways. Which one is better?
In the end we want a source image ( the .mov file) that has the most amount of "angular resolution" possible. Angular resolution is "pixels per degree"
Ok try to grasp this definition for a moment.
Panoramic resolution is the angular pixel density expressed in units of pixels per degree as determined by the focal length.
Panoramic resolution = the quality of our pano as it appears to us looking at.
Angular resolution = pixels per degree. Meaning as FOV changes we will have more or less pixels per "degree". As in 90 degrees. Also referred to as "quality of the image"
Focal length = The "filed of view" set as the default or starting FOV
If you are feeling spunky you can read up on this and play with the calculators found here
This means the "angular resolution" will change depending on the display size and most importantly the "filed of view" you set for the default or what the pano starts at.
This starting FOV is normally 90 deg. This is what the human eye sees clearly not counting the very edges of your peripheral vision. This is why a FOV of 90 deg in a pano appears to have the least distortion. If you increase FOV ( zoom out) the image is distorted BUT quality increases(sharpness). If you decrease the FOV (zoom in) the image will become flatter and at some point you will begin to see pixels - quality is
Hi Gen lee
Thank you for your great reply.
If I am understanding it correctly, seizing down is normally equal to loosing quality. We try to prevent loosing quality to seizing down in steps and harpen each time.
My next question is then. If I only need a pano at 450x350 pixels, would it not be better to stitch it at 2000x1000?? Then I do not have to size down and I can avoid loosing quality.
- a newbie trying to improve
- a newbie trying to improve