Restoration of the Parque Forestal Image

Translation (c) 2004 Scott Corrales


INEXPLICATA
The Journal of Hispanic Ufology
July 18, 2004

Source: CIFAE - CHILE

Date: 06.17.04

Restoration of the Parque Forestal Image

In regard to the analysis of the Parque Forestal photograph, CIFAE delivers this interesting report by our scientific director to the international UFO research community.

In this work we shall present a brief introduction to the restoration of images by means of deconvolution, which is the only correct technique for eliminating certain defects from images. We also attempt to restore the Parque Forestal image, although in this case it isn't possible to obtain very good results. We are not trying to debate what the subject of the photo is or is not, nor if it is of interest to ufology. The main subject under discussion here (restoration) should be of interest to any researcher who works with images.

An image distorted by being out of focus, astigmatism, "motion blur", atmospheric disturbances, etc. is mathematically represented as the convolution between the non-defective image and a PSF (point spread function). PSF is a function that can be interpreted as the distorted image of a white spot on a black background, and each image can have a different PSF depending on the camera and the conditions under which the image was acquired.

In order to restore the image and to eliminate these defects, one must perform the reverse operation, in other words, deconvolute the distorted image against the corresponding PSF. Deconvolution isn't hard to achieve, but the deconvolution algorithms tend to be sensible to the image's "noise" (fluctuations) and iterative methods can easily diverge. The worst problem arises when the PSF isn't known a priori, and in this case it is necessary to develop some method that can extract the PSF from the defective image, what is known as blind deconvolution. This problem can be hard to deal with, and may be considered as one of the great challenges of the field of image analysis.

For example, Figure 1 shows a restoration made by Michael Cannon in the 1970s, and who determined PSF based on a cepstral analysis and later deconvoluted the image using Wiener's Filter

Figure 1: An example of blind deconvolution.

It is also interesting to note the recent work of Alfred Carasso, who has achieved very impressive results, although the applicability of his methods is limited.

It is hard to find commercial software capable of performing blind deconvolutions. Some programs like AutoDeblur have been specifically developed for microscopy. The Unshake program is a general purpose one, but is based on a very simple method that doesn't always work well. Other software packages like Iris are only useful in astrophotography. The majority of the available programs require a priori knowledge of PSF, such as Unblur , which is general purpose. FocusFixer is only useful in dealing with out-of-focus image and Focus Magic corrects simple out of focus and motion blur.

For the time being we do not have access to a better commercial software package, therefore the Parque Forestal image was restored searching for PSF by simpler means. Knowing the basics of deconvolution, it is possible to know something about PSF by simple inspection of the photo. One can see that it appears in motion blur at angle of some 150 degrees. Using the Fix Motion Blur filter in Focus Magic at this angle, the distance could be established through trial and error, and turned out to be 11 pixels. In other words, the PSF that accoutns for the motion blur would be a straight line with a length of 11 pixels at an angle of 150 degrees. Upon eliminating the motion blur, it can be seen that the image is folded in the same direction (150 degrees) and at a distance of 17 pixels. In other words, the PSF that acknowledges this unfolding is two points at a distance of 17 pixels and at an angle of 150 degrees. The PSF was built using this inforamtion and the image was deconvoluted using Unblur. Focus Magic could not correct the unfolding. The result is shown in Figure 2, where it is noted that the image is partly restored, but with an increase in the amount of "noise". Contrast was not augmented artificially. The original photo has a lower contrast because the colors and shapes are mixed when an image is hazy, and upon restoring it, the colors and details of the real scene are also recovered.

According to expert James Caron, this image has several difficutlites that do not allow for its full restoration:

Due to poor lighting, a weak signal was obtained with excessive motion blur due to the long exposure time (0.1 seg).

The object of interest is moving and has a different, more sever motion blur than the static objects in the scene.

The image is compressed, which truncates the information necessary for its restoration.

In other words, the photo is compressed in the JPG format, which eliminates part of the information contained in the original image in order to reduce the storage size. The lack of lighting brings as a consequence an excessive amount of noise, especially in the restored image. This implies that there isn't a single PSF for the whole photo, which makes restoration difficult. On the other hand, any evidence of motion may be useful in intepreting the image. In our restoration, at least, some details and colors can be distinguished, and from the photo analysis point of view, it is the best that could be done in this case.

Figure 2: Restoration of the Parque Forestal photo.

Regarding the possibility of fraud in this photo, it would be easy to dot. The figure could have been pasted on the photo and the resulting borders could ahve been erased through motion blur. On the other hand, the photo's EXIF could have been edited in order to simulate an original photo. But the fact that it is easy to forge does not constitute evidence of a hoax, although it could provide good grounds for ignoring this case or not taking it seriously. We have hitherto not received any valid arguments supporting the hoax hypothesis. Some have said that some lines or squares of well-defined borders can be noted, but this is only a defect of the JPG format.

In closing, something should be said about other image processing techniques. Some effects that highlight the image's details are very common, such as contrast and saturation, sharpening, unsharp mask, border detection, emboss, solarizing, etc. But it is obvious that none of these filters would be capable of restoring an image such as the one in Figure 1. many of these filters are deceptive because they alter the image, giving the false impression of having improved it, but they are at times only useful for aesthetic purposes. It is advisable not to employ a filter if the basics of its operation are unknown.

As an example of a poor job, we mention the one created with a fantasy filter at the request of the IIEE in Figure 3 A, which could be interpreted as a dog. The idea is not to argue if the subject is a dog or not; all that is shown here is that this particular work is poorly executed. The diagonal lines that betray the presence of motion blur are highly visible, and one of the legs is unfolded in the same way as in the original photo. Overall, this image has the same geometry as the original, but different coloring. In other words, there is no evidence of restoration in this retouching. The way in which the colors have been altered is also noticeable.

Figure 3: A-Retouching performed for IIEE. B and C - Sculpture 3D

Investigation conducted by physicist Andres Duarte.

Scientific Director of CIFAE Chile

-----

A Report from KODAK Chile

The Kodak report, while not certifying the process of "acknowlegement of objects or external elements within a digital photo" by company policy, has been an important source document in this reseach.

Some parts of this report contain the following information:

The information confirms that the photo was indeed taken using a Kodak DX 6490 camera.

Dimension 1656x1242 pixels
Color depth 24 bpp
Density 96x96 dpi
Subsampling 2:1:1
Color space YCbCr
Mode Baseline
Compression JPEG
Exposure time 0.1 sec
F-Number 3.6
Exposure prog Normal program
EXIF ver 2.2
Time original 2004:05:10 17:40:18
Time digitized 2004:05:10 17:40:18
Component conf YCbCr
Shutter speed 197121 s
Aperture 3.6
Exposure bias 3.6
Max aperture 3.6
Metering mode Pattern
Light source Unknown Flash
Focal length 52.6 mm
Maker note KDK0101IDX6490
FlashPix ver 1.0
Colorspace sRGB
XY Dimension 1656x1242 pixels
Exposure index 140
Sensing method One- chip color area sensor
File source DSC
Scene source Directly photographed image

Sincerely,

Kodak Chile

-----

The CIFAE Report

Among some of the conclusions that can be drawn from the CIFAE report, we can say that the procedures applied up to now in analyzing the image of Parque Forestal, and in which some have thought to see a dog, are mistaken. At least in the positions know.

In other words, if this image in fact corresponds to a canine, the superficial analyses performed until now have NOT proved it.

CIFAE will continue to analyse the image and we shall continue to keep our readers abreast of further developments.

-----

[UFOINFO Note: See http://www.ovniaventura.com/foto_parque.htm for the original report with images.

Translation (c) 2004 Scott Corrales IHU
Special thanks to CIFAE

UFOINFO would like to thank Scott Corrales (Inexplicata) and UFO UpDates for granting permission to use this article. To keep up to date with follow-up reports and discussions you are advised to subscribe to UFO UpDates by contacting Errol Bruce-Knapp at: http://ufoupdateslist.com


UFOINFO
Copyright © UFOINFO
Articles are Copyright of the Author or Compiler