Did you spectrally subset the Hyperion data? Typically Hyperion has more than 112 "good" spectral bands. It would be unusual that the first PCA contains more than 99% of the information but I suppose it might matter what sort of image you have. Perhaps a large water scene or other very homogeneous scene would show results like that.

Yes, I spectrally subset my Hyperion dataset and of the 242 a subset of 112 bands spectrally subsetted from the Hyperion data set. I should not have done this step first? Thank you very much for your reply!

No, I think it is fine if you subset out the bad bands in Hyperion as there are quite a few. I just having seen it return only one PC with all of the variability (99.5%). Is your scene homogeneous?

I am sorry but I can't understand! What do you mean homogenous?

I also have PC2, PC3 and PC4...

Is there a diversity of materials in your image or only one or two materials? For example, a field of the same crop where you can only see plants or a large body of water would show very few materials. In this case, you may get all the relevant information in very few PC bands as in your case. Can you post the statistics from your eigenvalues?

I masked the Hyperion image, in order to avoid the vegetation and the water, because I investigate the rock formations... Yes, the statistics from the eigenvalues are the above:

Eigenvalues
Num 1 0.445516
Num 2 0.001159
Num 3 0.000465
Num 4 0.000210
Num 5 0.000073
Num 6 0.000035
Num 7 0.000022
Num 8 0.000019
Num 9 0.000014
Num 10 0.000008
Num 11 0.000007
Num 12 0.000006
Num 13 0.000005
Num 14 0.000005
Num 15 0.000005
Num 16 0.000004
Num 17 0.000004
Num 18 0.000004
Num 19 0.000004
Num 20 0.000004
Num 21 0.000003
Num 22 0.000003
Num 23 0.000003
Num 24 0.000003
Num 25 0.000003
Num 26 0.000003
Num 27 0.000003
Num 28 0.000002
Num 29 0.000002
Num 30 0.000002
Num 31 0.000002
Num 32 0.000002
Num 33 0.000002
Num 34 0.000002
Num 35 0.000002
Num 36 0.000002
Num 37 0.000002
Num 38 0.000002
Num 39 0.000001
Num 40 0.000001
Num 41 0.000001
Num 42 0.000001
Num 43 0.000001
Num 44 0.000001
Num 45 0.000001
Num 46 0.000001
Num 47 0.000001
Num 48 0.000001
Num 49 0.000001
Num 50 0.000001
Num 51 0.000001
Num 52 0.000001
Num 53 0.000001
Num 54 0.000001
Num 55 0.000001
Num 56 0.000001
Num 57 0.000001
Num 58 0.000001
Num 59 0.000001
Num 60 0.000001
Num 61 0.000001
Num 62 0.000001
Num 63 0.000001
Num 64 0.000001
Num 65 0.000001
Num 66 0.000001
Num 67 0.000000
Num 68 0.000000
Num 69 0.000000
Num 70 0.000000
Num 71 0.000000
Num 72 0.000000
Num 73 0.000000
Num 74 0.000000
Num 75 0.000000
Num 76 0.000000
Num 77 0.000000
Num 78 0.000000
Num 79 0.000000
Num 80 0.000000
Num 81 0.000000
Num 82 0.000000
Num 83 0.000000
Num 84 0.000000
Num 85 0.000000
Num 86 0.000000
Num 87 0.000000
Num 88 0.000000
Num 89 0.000000
Num 90 0.000000
Num 91 0.000000
Num 92 0.000000
Num 93 0.000000
Num 94 0.000000
Num 95 0.000000
Num 96 0.000000
Num 97 0.000000
Num 98 0.000000
Num 99 0.000000
Num 100 0.000000
Num 101 0.000000
Num 102 0.000000
Num 103 0.000000
Num 104 0.000000
Num 105 0.000000
Num 106 0.000000
Num 107 0.000000
Num 108 0.000000
Num 109 0.000000
Num 110 0.000000
Num 111 0.000000
Num 112 0.000000

I ran a PCA on the tutorial data set, HyperionForest.dat, and it includes all the original bands with some set to bad bands. The majority of the variability was contained in the first 5 bands while the rest included a lot of noise. Is this what you see in your data? Perhaps it is normal for Hyperion which is known to be a very noisy sensor.